ELF>� @ah@8@�Zh�Zh �]h�]��]�`h �]h�]��]��$$P�tdXhXhXh��Q�tdR�td�]h�]��]�HHGNU�, g�le��1�n� �E h �@#% ��g^g^�g^!g^#g^��qs| Dx㚭�(j;g^g^�g^�g^�g^�g^�g^�g^�g^ U , F"'��p+��T�}�i� �� P �`�� 50^@�__gmon_start___ITM_deregisterTMCloneTable_ITM_registerTMCloneTable__cxa_finalize_Jv_RegisterClasses__nvrtc_builtins__12_1_getBuiltinHeader__nvrtc_builtins__12_1_getArchBuiltins50__nvrtc_builtins__12_1_getArchBuiltins52__nvrtc_builtins__12_1_getArchBuiltins53__nvrtc_builtins__12_1_getArchBuiltins60__nvrtc_builtins__12_1_getArchBuiltins61__nvrtc_builtins__12_1_getArchBuiltins62__nvrtc_builtins__12_1_getArchBuiltins70__nvrtc_builtins__12_1_getArchBuiltins72__nvrtc_builtins__12_1_getArchBuiltins75__nvrtc_builtins__12_1_getArchBuiltins80__nvrtc_builtins__12_1_getArchBuiltins86__nvrtc_builtins__12_1_getArchBuiltins87__nvrtc_builtins__12_1_getArchBuiltins89__nvrtc_builtins__12_1_getArchBuiltins90__nvrtc_builtins__12_1_getArchBuiltinslibstdc++.so.6libm.so.6libgcc_s.so.1libc.so.6libnvrtc-builtins.so.12.1GLIBC_2.2.5a�(a�'ui A�]�� ]�` �]��]��_��_��_��_��_�H��H�EV�H��t�#H��5RV��%TV�@�%V�f��%*V�f�H�=AV�H�AV�UH)�H��H��vH��U�H��t ]��fD]�@f.�H�=V�H�5�U�UH)�H��H��H��H��?H�H��tH��U�H��t]��f�]�@f.��=�U�u'H�=�U�UH��tH�=RS��E��H��]��U��@f.�H�=!S�H�?u�^��fDH�!U�H��t�UH��]�@��H��)IH��ÐH��)IH��ÐH��&H�B,IÐH��&H�2SKÐH��&H�"zMÐH�XOH��OÐH�XOH�b�QÐH�XOH��?TÐH��2H��VÐH��2H��XÐH��2H�"�ZÐH��8H��&]ÐH��8H�B__ÐH��8H�җaÐH��8H�b�cÐH�@CH��fÐ��2t[��4t^��5ta��<td��=tg��>tj��Ftm��Htp��Kts��Ptv��Vty��Wt|��Yt��Zu ��DH�1��fD��f.��2t[��4t^��5ta��<td��=tg��>tj��Ftm��Htp��Kts��Ptv��Vty��Wt|��Yt��Zu ��DH�1��fD��#��+��3��;��C��K��H��H��/* * Copyright 1993-2016 NVIDIA Corporation. All rights reserved. * * NOTICE TO LICENSEE: * * This source code and/or documentation ("Licensed Deliverables") are * subject to NVIDIA intellectual property rights under U.S. and * international Copyright laws. * * These Licensed Deliverables contained herein is PROPRIETARY and * CONFIDENTIAL to NVIDIA and is being provided under the terms and * conditions of a form of NVIDIA software license agreement by and * between NVIDIA and Licensee ("License Agreement") or electronically * accepted by Licensee. Notwithstanding any terms or conditions to * the contrary in the License Agreement, reproduction or disclosure * of the Licensed Deliverables to any third party without the express * written consent of NVIDIA is prohibited. * * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE. IT IS * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE * OF THESE LICENSED DELIVERABLES. * * U.S. Government End Users. These Licensed Deliverables are a * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT * 1995), consisting of "commercial computer software" and "commercial * computer software documentation" as such terms are used in 48 * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government * only as a commercial end item. Consistent with 48 C.F.R.12.212 and * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all * U.S. Government End Users acquire the Licensed Deliverables with * only those rights set forth herein. * * Any use of the Licensed Deliverables in individual and commercial * software must include, in the user documentation and internal * comments to the code, the above Disclaimer and U.S. Government End * Users Notice. */ #define __CUDACC__ 1 #define __CUDANVVM__ 1 #if !defined(__cplusplus) #error "unexpected!" #endif #if defined(__LP64__) typedef unsigned long size_t; typedef long int ptrdiff_t; #else typedef unsigned long long size_t; typedef long long int ptrdiff_t; #endif typedef long clock_t; #ifndef NULL #define NULL 0 #endif #if defined(__cplusplus) extern "C" { #endif __attribute__((device)) int __nvvm_reflect(const char*); #if defined(__cplusplus) } #endif #if !defined(__CUDA_RUNTIME_H__) #define __CUDA_RUNTIME_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_H__ #endif #if !defined(__CUDACC_RTC__) #if defined(__GNUC__) #if defined(__clang__) || (!defined(__PGIC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) #pragma GCC diagnostic push #endif #if defined(__clang__) || (!defined(__PGIC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 2))) #pragma GCC diagnostic ignored "-Wunused-function" #endif #elif defined(_MSC_VER) #pragma warning(push) #pragma warning(disable: 4820) #endif #endif #ifdef __QNX__ #if (__GNUC__ == 4 && __GNUC_MINOR__ >= 7) typedef unsigned size_t; #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_config.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_config.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_CONFIG_H__ #endif #if !defined(__HOST_CONFIG_H__) #define __HOST_CONFIG_H__ #if defined(__CUDACC__) #if defined(__CUDACC_RTC__) #define _CRTIMP #define __THROW #else #if !defined(__GNUC__) && !defined(_WIN32) #error --- !!! UNSUPPORTED COMPILER !!! --- #endif #if defined(__PGIC__) #if !defined(__GNUC__) || !defined(__LP64__) || !defined(__linux__) #error -- unsupported pgc++ configuration! pgc++ is supported only on Linux x86_64! #endif #endif #if defined(__powerpc__) #if !defined(__powerpc64__) || !defined(__LITTLE_ENDIAN__) #error -- unsupported PPC platform! Only 64-bit little endian PPC is supported! #endif #endif #if defined(__APPLE__) && defined(__MACH__) && !defined(__clang__) #error -- clang and clang++ are the only supported host compilers on Mac OS X! #endif #if !__NV_NO_HOST_COMPILER_CHECK #if defined(__ICC) #if (__ICC != 1500 && __ICC != 1600 && __ICC != 1700 && __ICC != 1800 && !(__ICC >= 1900 && __ICC <= 2021)) || !defined(__GNUC__) || !defined(__LP64__) #error -- unsupported ICC configuration! Only ICC 15.0, ICC 16.0, ICC 17.0, ICC 18.0, ICC 19.x and 20.x on Linux x86_64 are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. #endif #endif #if defined(__powerpc__) #if defined(__ibmxl_vrm__) && !(__ibmxl_vrm__ >= 0x0d010000 && __ibmxl_vrm__ < 0x0d020000) && \ !(__ibmxl_vrm__ >= 0x10010000 && __ibmxl_vrm__ < 0x10020000) #error -- unsupported xlC version! only xlC 13.1 and 16.1 are supported. The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. #endif #endif #if defined(__GNUC__) #if __GNUC__ > 12 #error -- unsupported GNU version! gcc versions later than 12 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. #endif #if defined(__clang__) && !defined(__ibmxl_vrm__) && !defined(__ICC) && !defined(__HORIZON__) && !defined(__APPLE__) #if (__clang_major__ >= 16) || (__clang_major__ < 3) || ((__clang_major__ == 3) && (__clang_minor__ < 3)) #error -- unsupported clang version! clang version must be less than 16 and greater than 3.2 . The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. #endif #endif #endif #if defined(_WIN32) #if _MSC_VER < 1910 || _MSC_VER >= 1940 #error -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. #elif _MSC_VER >= 1910 && _MSC_VER < 1910 #pragma message("support for this version of Microsoft Visual Studio has been deprecated! Only the versions between 2017 and 2022 (inclusive) are supported!") #endif #endif #endif #if defined(__APPLE__) #define _CRTIMP #define _ACRTIMP #define __THROW #if defined(__BLOCKS__) #undef __BLOCKS__ #endif #elif defined(__ANDROID__) #define _CRTIMP #define _ACRTIMP #define __THROW #elif defined(__QNX__) #define _CRTIMP #define _ACRTIMP #define __THROW #elif defined(__HORIZON__) #define _CRTIMP #define _ACRTIMP #define __THROW #elif defined(__GNUC__) #define _CRTIMP #define _ACRTIMP #include <features.h> #elif defined(_WIN32) #if _MSC_VER >= 1500 #undef _USE_DECLSPECS_FOR_SAL #define _USE_DECLSPECS_FOR_SAL \ 1 #endif #if !defined(_CRT_NONSTDC_NO_WARNINGS) #define _CRT_NONSTDC_NO_WARNINGS #endif #if !defined(_CRT_SECURE_NO_WARNINGS) #define _CRT_SECURE_NO_WARNINGS #endif #if !defined(NOMINMAX) #define NOMINMAX #endif #include <crtdefs.h> #if _MSC_VER >= 1900 #include <corecrt.h> #endif #define __THROW #endif #endif #if defined(__cplusplus) && defined(__CUDA_ARCH__) && (defined(__PGIC__) || defined(__CUDACC_RTC__) || (defined(_WIN32) && defined(_MSC_VER))) #if __CUDACC_RTC__ typedef char *va_list; #else #include <cstdarg> #endif #undef va_start #undef va_end #undef va_arg #ifdef __PGIC__ #undef __builtin_va_end #define va_start(v,l) __builtin_alt_va_start(v,l) #define va_end(v) __builtin_va_end(v) #define va_arg(v,l) __builtin_alt_va_arg(v,l) #if (__cplusplus >= 201103L) #undef va_copy #define va_copy(d,s) __builtin_va_copy(d,s) #endif #else #define va_start(ap, x) (__cu_va_start(&ap, x)) #define va_end(ap) (__cu_va_end(&ap)) #define va_arg(ap, t) (*((t *)__cu_va_arg(&ap, (t *)0))) #if (_MSC_VER >= 1800) || (defined(__CUDACC_RTC__) && (__cplusplus >= 201103L)) #undef va_copy #define va_copy(apd, aps) (__cu_va_copy(&(apd), &(aps))) #endif #endif #endif #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_CONFIG_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_CONFIG_H__ #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__LIBRARY_TYPES_H__) #define __LIBRARY_TYPES_H__ typedef enum cudaDataType_t { CUDA_R_16F = 2, CUDA_C_16F = 6, CUDA_R_16BF = 14, CUDA_C_16BF = 15, CUDA_R_32F = 0, CUDA_C_32F = 4, CUDA_R_64F = 1, CUDA_C_64F = 5, CUDA_R_4I = 16, CUDA_C_4I = 17, CUDA_R_4U = 18, CUDA_C_4U = 19, CUDA_R_8I = 3, CUDA_C_8I = 7, CUDA_R_8U = 8, CUDA_C_8U = 9, CUDA_R_16I = 20, CUDA_C_16I = 21, CUDA_R_16U = 22, CUDA_C_16U = 23, CUDA_R_32I = 10, CUDA_C_32I = 11, CUDA_R_32U = 12, CUDA_C_32U = 13, CUDA_R_64I = 24, CUDA_C_64I = 25, CUDA_R_64U = 26, CUDA_C_64U = 27, CUDA_R_8F_E4M3 = 28, CUDA_R_8F_E5M2 = 29, } cudaDataType; typedef enum libraryPropertyType_t { MAJOR_VERSION, MINOR_VERSION, PATCH_LEVEL } libraryPropertyType; #ifndef __cplusplus typedef enum cudaDataType_t cudaDataType_t; typedef enum libraryPropertyType_t libraryPropertyType_t; #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #ifdef __CUDACC_RTC__ #ifndef __NV_TARGET_H #define __NV_TARGET_H #if defined(__NVCC__) || defined(__CUDACC_RTC__) # define _NV_COMPILER_NVCC #elif defined(__NVCOMPILER) && __cplusplus >= 201103L # define _NV_COMPILER_NVCXX #elif defined(__clang__) && defined(__CUDA__) && defined(__CUDA_ARCH__) # define _NV_COMPILER_CLANG_CUDA #endif #if defined(__CUDACC_RTC__) # define _NV_FUNCTION_ANNOTATION __device__ #else # define _NV_FUNCTION_ANNOTATION #endif #if defined(_NV_COMPILER_NVCXX) # define _NV_BITSET_ATTRIBUTE [[nv::__target_bitset]] #else # define _NV_BITSET_ATTRIBUTE #endif #if (!defined(__ibmxl__)) && \ ((defined(__cplusplus) && __cplusplus >= 201103L) || \ (defined(_MSC_VER) && _MSVC_LANG >= 201103L)) # define _NV_TARGET_CPP11 #endif #if defined(_NV_TARGET_CPP11) namespace nv { namespace target { namespace detail { typedef unsigned long long base_int_t; constexpr base_int_t all_hosts = 1; constexpr base_int_t sm_35_bit = 1 << 1; constexpr base_int_t sm_37_bit = 1 << 2; constexpr base_int_t sm_50_bit = 1 << 3; constexpr base_int_t sm_52_bit = 1 << 4; constexpr base_int_t sm_53_bit = 1 << 5; constexpr base_int_t sm_60_bit = 1 << 6; constexpr base_int_t sm_61_bit = 1 << 7; constexpr base_int_t sm_62_bit = 1 << 8; constexpr base_int_t sm_70_bit = 1 << 9; constexpr base_int_t sm_72_bit = 1 << 10; constexpr base_int_t sm_75_bit = 1 << 11; constexpr base_int_t sm_80_bit = 1 << 12; constexpr base_int_t sm_86_bit = 1 << 13; constexpr base_int_t sm_87_bit = 1 << 14; constexpr base_int_t sm_89_bit = 1 << 15; constexpr base_int_t sm_90_bit = 1 << 16; constexpr base_int_t all_devices = sm_35_bit | sm_37_bit | sm_50_bit | sm_52_bit | sm_53_bit | sm_60_bit | sm_61_bit | sm_62_bit | sm_70_bit | sm_72_bit | sm_75_bit | sm_80_bit | sm_86_bit | sm_87_bit | sm_89_bit | sm_90_bit; struct _NV_BITSET_ATTRIBUTE target_description { base_int_t targets; _NV_FUNCTION_ANNOTATION constexpr target_description(base_int_t a) : targets(a) { } }; enum class sm_selector : base_int_t { sm_35 = 35, sm_37 = 37, sm_50 = 50, sm_52 = 52, sm_53 = 53, sm_60 = 60, sm_61 = 61, sm_62 = 62, sm_70 = 70, sm_72 = 72, sm_75 = 75, sm_80 = 80, sm_86 = 86, sm_87 = 87, sm_89 = 89, sm_90 = 90, }; _NV_FUNCTION_ANNOTATION constexpr base_int_t toint(sm_selector a) { return static_cast<base_int_t>(a); } _NV_FUNCTION_ANNOTATION constexpr base_int_t bitexact(sm_selector a) { return toint(a) == 35 ? sm_35_bit : toint(a) == 37 ? sm_37_bit : toint(a) == 50 ? sm_50_bit : toint(a) == 52 ? sm_52_bit : toint(a) == 53 ? sm_53_bit : toint(a) == 60 ? sm_60_bit : toint(a) == 61 ? sm_61_bit : toint(a) == 62 ? sm_62_bit : toint(a) == 70 ? sm_70_bit : toint(a) == 72 ? sm_72_bit : toint(a) == 75 ? sm_75_bit : toint(a) == 80 ? sm_80_bit : toint(a) == 86 ? sm_86_bit : toint(a) == 87 ? sm_87_bit : toint(a) == 89 ? sm_89_bit : toint(a) == 90 ? sm_90_bit : 0; } _NV_FUNCTION_ANNOTATION constexpr base_int_t bitrounddown(sm_selector a) { return toint(a) >= 90 ? sm_90_bit : toint(a) >= 89 ? sm_89_bit : toint(a) >= 87 ? sm_87_bit : toint(a) >= 86 ? sm_86_bit : toint(a) >= 80 ? sm_80_bit : toint(a) >= 75 ? sm_75_bit : toint(a) >= 72 ? sm_72_bit : toint(a) >= 70 ? sm_70_bit : toint(a) >= 62 ? sm_62_bit : toint(a) >= 61 ? sm_61_bit : toint(a) >= 60 ? sm_60_bit : toint(a) >= 53 ? sm_53_bit : toint(a) >= 52 ? sm_52_bit : toint(a) >= 50 ? sm_50_bit : toint(a) >= 37 ? sm_37_bit : toint(a) >= 35 ? sm_35_bit : 0; } _NV_FUNCTION_ANNOTATION constexpr target_description is_exactly(sm_selector a) { return target_description(bitexact(a)); } _NV_FUNCTION_ANNOTATION constexpr target_description provides(sm_selector a) { return target_description(~(bitrounddown(a) - 1) & all_devices); } _NV_FUNCTION_ANNOTATION constexpr target_description operator&&(target_description a, target_description b) { return target_description(a.targets & b.targets); } _NV_FUNCTION_ANNOTATION constexpr target_description operator||(target_description a, target_description b) { return target_description(a.targets | b.targets); } _NV_FUNCTION_ANNOTATION constexpr target_description operator!(target_description a) { return target_description(~a.targets & (all_devices | all_hosts)); } } using detail::target_description; using detail::sm_selector; constexpr target_description is_host = target_description(detail::all_hosts); constexpr target_description is_device = target_description(detail::all_devices); constexpr target_description any_target = target_description(detail::all_hosts | detail::all_devices); constexpr target_description no_target = target_description(0); constexpr sm_selector sm_35 = sm_selector::sm_35; constexpr sm_selector sm_37 = sm_selector::sm_37; constexpr sm_selector sm_50 = sm_selector::sm_50; constexpr sm_selector sm_52 = sm_selector::sm_52; constexpr sm_selector sm_53 = sm_selector::sm_53; constexpr sm_selector sm_60 = sm_selector::sm_60; constexpr sm_selector sm_61 = sm_selector::sm_61; constexpr sm_selector sm_62 = sm_selector::sm_62; constexpr sm_selector sm_70 = sm_selector::sm_70; constexpr sm_selector sm_72 = sm_selector::sm_72; constexpr sm_selector sm_75 = sm_selector::sm_75; constexpr sm_selector sm_80 = sm_selector::sm_80; constexpr sm_selector sm_86 = sm_selector::sm_86; constexpr sm_selector sm_87 = sm_selector::sm_87; constexpr sm_selector sm_89 = sm_selector::sm_89; constexpr sm_selector sm_90 = sm_selector::sm_90; using detail::is_exactly; using detail::provides; } } #endif #ifndef _NV__TARGET_MACROS #define _NV__TARGET_MACROS #if defined(__GNUC__) #pragma GCC system_header #endif #if defined(_NV_TARGET_CPP11) # define _NV_EVAL1(...) __VA_ARGS__ # define _NV_EVAL(...) _NV_EVAL1(__VA_ARGS__) #else # define _NV_EVAL1(x) x # define _NV_EVAL(x) _NV_EVAL1(x) #endif #define _NV_CONCAT_EVAL1(l, r) _NV_EVAL(l ## r) #define _NV_CONCAT_EVAL(l, r) _NV_CONCAT_EVAL1(l, r) #define _NV_IF_0(t, f) f #define _NV_IF_1(t, f) t #define _NV_IF_BIT(b) _NV_EVAL(_NV_IF_##b) #define _NV_IF__EVAL(fn, t, f) _NV_EVAL(fn(t, f)) #define _NV_IF_EVAL(cond, t, f) _NV_IF__EVAL(_NV_IF_BIT(cond), t, f) #define _NV_IF1(cond, t, f) _NV_IF_EVAL(cond, t, f) #define _NV_IF(cond, t, f) _NV_IF1(_NV_EVAL(cond), _NV_EVAL(t), _NV_EVAL(f)) #if defined(_NV_TARGET_CPP11) #define _NV_ARG32(...) _NV_EVAL(_NV_ARG32_0(__VA_ARGS__)) #define _NV_ARG32_0( \ _0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15, \ _16, _17, _18, _19, _20, _21, _22, _23, _24, _25, _26, _27, _28, _29, _30, _31, ...) _31 #define _NV_HAS_COMMA(...) _NV_ARG32(__VA_ARGS__, \ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, \ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0) #define _NV_TRIGGER_PARENTHESIS_(...) , #define _NV_ISEMPTY(...) \ _NV_ISEMPTY0( \ \ _NV_EVAL(_NV_HAS_COMMA(__VA_ARGS__)), \ \ _NV_EVAL(_NV_HAS_COMMA(_NV_TRIGGER_PARENTHESIS_ __VA_ARGS__)), \ \ _NV_EVAL(_NV_HAS_COMMA(__VA_ARGS__ ( ))), \ \ _NV_EVAL(_NV_HAS_COMMA(_NV_TRIGGER_PARENTHESIS_ __VA_ARGS__ ( ))) \ ) #define _NV_PASTE5(_0, _1, _2, _3, _4) _0 ## _1 ## _2 ## _3 ## _4 #define _NV_ISEMPTY0(_0, _1, _2, _3) _NV_HAS_COMMA(_NV_PASTE5(_NV_IS_EMPTY_CASE_, _0, _1, _2, _3)) #define _NV_IS_EMPTY_CASE_0001 , #define _NV_REMOVE_PAREN(...) _NV_REMOVE_PAREN1(__VA_ARGS__) #define _NV_REMOVE_PAREN1(...) _NV_STRIP_PAREN(_NV_IF(_NV_TEST_PAREN(__VA_ARGS__), (_NV_STRIP_PAREN(__VA_ARGS__)), (__VA_ARGS__))) #define _NV_STRIP_PAREN2(...) __VA_ARGS__ #define _NV_STRIP_PAREN1(...) _NV_STRIP_PAREN2 __VA_ARGS__ #define _NV_STRIP_PAREN(...) _NV_STRIP_PAREN1(__VA_ARGS__) #define _NV_TEST_PAREN(...) _NV_TEST_PAREN1(__VA_ARGS__) #define _NV_TEST_PAREN1(...) _NV_TEST_PAREN2(_NV_TEST_PAREN_DUMMY __VA_ARGS__) #define _NV_TEST_PAREN2(...) _NV_TEST_PAREN3(_NV_CONCAT_EVAL(_, __VA_ARGS__)) #define _NV_TEST_PAREN3(...) _NV_EVAL(_NV_FIRST_ARG(__VA_ARGS__)) #define __NV_PAREN_YES 1 #define __NV_PAREN_NO 0 #define _NV_TEST_PAREN_DUMMY(...) _NV_PAREN_YES #define __NV_TEST_PAREN_DUMMY __NV_PAREN_NO, #define _NV_FIRST_ARG1(x, ...) x #define _NV_FIRST_ARG(x, ...) _NV_FIRST_ARG1(x) #define _NV_REMOVE_FIRST_ARGS1(...) __VA_ARGS__ #define _NV_REMOVE_FIRST_ARGS(x, ...) _NV_REMOVE_FIRST_ARGS1(__VA_ARGS__) #define _NV_NUM_ARGS(...) _NV_NUM_ARGS0(__VA_ARGS__) #define _NV_NUM_ARGS0(...) _NV_EVAL(_NV_NUM_ARGS1(__VA_ARGS__)) #define _NV_NUM_ARGS1(...) _NV_IF(_NV_ISEMPTY(__VA_ARGS__), 0, _NV_NUM_ARGS2(__VA_ARGS__)) #define _NV_NUM_ARGS2(...) _NV_ARG32(__VA_ARGS__, \ 31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16, \ 15,14,13,12,11,10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0) #define _NV_DISPATCH_N_IMPL1(name, ...) _NV_EVAL(name(__VA_ARGS__)) #define _NV_DISPATCH_N_IMPL0(depth, name, ...) _NV_DISPATCH_N_IMPL1(_NV_CONCAT_EVAL(name, depth), __VA_ARGS__) #define _NV_DISPATCH_N_IMPL(name, ...) _NV_DISPATCH_N_IMPL0(_NV_NUM_ARGS(__VA_ARGS__), name, __VA_ARGS__) #define _NV_DISPATCH_N_ARY(name, ...) _NV_DISPATCH_N_IMPL(name, __VA_ARGS__) #endif #if defined(__GNUC__) #pragma GCC system_header #endif # define _NV_TARGET_ARCH_TO_SELECTOR_350 nv::target::sm_35 # define _NV_TARGET_ARCH_TO_SELECTOR_370 nv::target::sm_37 # define _NV_TARGET_ARCH_TO_SELECTOR_500 nv::target::sm_50 # define _NV_TARGET_ARCH_TO_SELECTOR_520 nv::target::sm_52 # define _NV_TARGET_ARCH_TO_SELECTOR_530 nv::target::sm_53 # define _NV_TARGET_ARCH_TO_SELECTOR_600 nv::target::sm_60 # define _NV_TARGET_ARCH_TO_SELECTOR_610 nv::target::sm_61 # define _NV_TARGET_ARCH_TO_SELECTOR_620 nv::target::sm_62 # define _NV_TARGET_ARCH_TO_SELECTOR_700 nv::target::sm_70 # define _NV_TARGET_ARCH_TO_SELECTOR_720 nv::target::sm_72 # define _NV_TARGET_ARCH_TO_SELECTOR_750 nv::target::sm_75 # define _NV_TARGET_ARCH_TO_SELECTOR_800 nv::target::sm_80 # define _NV_TARGET_ARCH_TO_SELECTOR_860 nv::target::sm_86 # define _NV_TARGET_ARCH_TO_SELECTOR_870 nv::target::sm_87 # define _NV_TARGET_ARCH_TO_SELECTOR_890 nv::target::sm_89 # define _NV_TARGET_ARCH_TO_SELECTOR_900 nv::target::sm_90 # define _NV_TARGET_ARCH_TO_SM_350 35 # define _NV_TARGET_ARCH_TO_SM_370 37 # define _NV_TARGET_ARCH_TO_SM_500 50 # define _NV_TARGET_ARCH_TO_SM_520 52 # define _NV_TARGET_ARCH_TO_SM_530 53 # define _NV_TARGET_ARCH_TO_SM_600 60 # define _NV_TARGET_ARCH_TO_SM_610 61 # define _NV_TARGET_ARCH_TO_SM_620 62 # define _NV_TARGET_ARCH_TO_SM_700 70 # define _NV_TARGET_ARCH_TO_SM_720 72 # define _NV_TARGET_ARCH_TO_SM_750 75 # define _NV_TARGET_ARCH_TO_SM_800 80 # define _NV_TARGET_ARCH_TO_SM_860 86 # define _NV_TARGET_ARCH_TO_SM_870 87 # define _NV_TARGET_ARCH_TO_SM_890 89 # define _NV_TARGET_ARCH_TO_SM_900 90 #if defined(_NV_COMPILER_NVCXX) && defined(_NVHPC_CUDA) # define _NV_TARGET_VAL_SM_35 nv::target::sm_35 # define _NV_TARGET_VAL_SM_37 nv::target::sm_37 # define _NV_TARGET_VAL_SM_50 nv::target::sm_50 # define _NV_TARGET_VAL_SM_52 nv::target::sm_52 # define _NV_TARGET_VAL_SM_53 nv::target::sm_53 # define _NV_TARGET_VAL_SM_60 nv::target::sm_60 # define _NV_TARGET_VAL_SM_61 nv::target::sm_61 # define _NV_TARGET_VAL_SM_62 nv::target::sm_62 # define _NV_TARGET_VAL_SM_70 nv::target::sm_70 # define _NV_TARGET_VAL_SM_72 nv::target::sm_72 # define _NV_TARGET_VAL_SM_75 nv::target::sm_75 # define _NV_TARGET_VAL_SM_80 nv::target::sm_80 # define _NV_TARGET_VAL_SM_86 nv::target::sm_86 # define _NV_TARGET_VAL_SM_87 nv::target::sm_87 # define _NV_TARGET_VAL_SM_89 nv::target::sm_89 # define _NV_TARGET_VAL_SM_90 nv::target::sm_90 # define _NV_TARGET___NV_IS_HOST nv::target::is_host # define _NV_TARGET___NV_IS_DEVICE nv::target::is_device # define _NV_TARGET___NV_ANY_TARGET (nv::target::any_target) # define _NV_TARGET___NV_NO_TARGET (nv::target::no_target) # if defined(NV_TARGET_SM_INTEGER_LIST) # define NV_TARGET_MINIMUM_SM_SELECTOR _NV_FIRST_ARG(NV_TARGET_SM_SELECTOR_LIST) # define NV_TARGET_MINIMUM_SM_INTEGER _NV_FIRST_ARG(NV_TARGET_SM_INTEGER_LIST) # define __CUDA_MINIMUM_ARCH__ _NV_CONCAT_EVAL(_NV_FIRST_ARG(NV_TARGET_SM_INTEGER_LIST), 0) # endif # define _NV_TARGET_PROVIDES(q) nv::target::provides(q) # define _NV_TARGET_IS_EXACTLY(q) nv::target::is_exactly(q) #elif defined(_NV_COMPILER_NVCC) || defined (_NV_COMPILER_CLANG_CUDA) # define _NV_TARGET_VAL_SM_35 350 # define _NV_TARGET_VAL_SM_37 370 # define _NV_TARGET_VAL_SM_50 500 # define _NV_TARGET_VAL_SM_52 520 # define _NV_TARGET_VAL_SM_53 530 # define _NV_TARGET_VAL_SM_60 600 # define _NV_TARGET_VAL_SM_61 610 # define _NV_TARGET_VAL_SM_62 620 # define _NV_TARGET_VAL_SM_70 700 # define _NV_TARGET_VAL_SM_72 720 # define _NV_TARGET_VAL_SM_75 750 # define _NV_TARGET_VAL_SM_80 800 # define _NV_TARGET_VAL_SM_86 860 # define _NV_TARGET_VAL_SM_87 870 # define _NV_TARGET_VAL_SM_89 890 # define _NV_TARGET_VAL_SM_90 900 # if defined(__CUDA_ARCH__) # define _NV_TARGET_VAL __CUDA_ARCH__ # define NV_TARGET_MINIMUM_SM_SELECTOR _NV_CONCAT_EVAL(_NV_TARGET_ARCH_TO_SELECTOR_, __CUDA_ARCH__) # define NV_TARGET_MINIMUM_SM_INTEGER _NV_CONCAT_EVAL(_NV_TARGET_ARCH_TO_SM_, __CUDA_ARCH__) # define __CUDA_MINIMUM_ARCH__ __CUDA_ARCH__ # endif # if defined(__CUDA_ARCH__) # define _NV_TARGET_IS_HOST 0 # define _NV_TARGET_IS_DEVICE 1 # else # define _NV_TARGET_IS_HOST 1 # define _NV_TARGET_IS_DEVICE 0 # endif # if defined(_NV_TARGET_VAL) # define _NV_DEVICE_CHECK(q) (q) # else # define _NV_DEVICE_CHECK(q) (0) # endif # define _NV_TARGET_PROVIDES(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL >= q) # define _NV_TARGET_IS_EXACTLY(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL == q) #else # define _NV_COMPILER_NVCC # define _NV_TARGET_VAL_SM_35 350 # define _NV_TARGET_VAL_SM_37 370 # define _NV_TARGET_VAL_SM_50 500 # define _NV_TARGET_VAL_SM_52 520 # define _NV_TARGET_VAL_SM_53 530 # define _NV_TARGET_VAL_SM_60 600 # define _NV_TARGET_VAL_SM_61 610 # define _NV_TARGET_VAL_SM_62 620 # define _NV_TARGET_VAL_SM_70 700 # define _NV_TARGET_VAL_SM_72 720 # define _NV_TARGET_VAL_SM_75 750 # define _NV_TARGET_VAL_SM_80 800 # define _NV_TARGET_VAL_SM_86 860 # define _NV_TARGET_VAL_SM_87 870 # define _NV_TARGET_VAL_SM_89 890 # define _NV_TARGET_VAL_SM_90 900 # define _NV_TARGET_VAL 0 # define _NV_TARGET_IS_HOST 1 # define _NV_TARGET_IS_DEVICE 0 # define _NV_DEVICE_CHECK(q) (false) # define _NV_TARGET_PROVIDES(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL >= q) # define _NV_TARGET_IS_EXACTLY(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL == q) #endif #define _NV_TARGET___NV_PROVIDES_SM_35 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_35)) #define _NV_TARGET___NV_PROVIDES_SM_37 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_37)) #define _NV_TARGET___NV_PROVIDES_SM_50 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_50)) #define _NV_TARGET___NV_PROVIDES_SM_52 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_52)) #define _NV_TARGET___NV_PROVIDES_SM_53 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_53)) #define _NV_TARGET___NV_PROVIDES_SM_60 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_60)) #define _NV_TARGET___NV_PROVIDES_SM_61 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_61)) #define _NV_TARGET___NV_PROVIDES_SM_62 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_62)) #define _NV_TARGET___NV_PROVIDES_SM_70 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_70)) #define _NV_TARGET___NV_PROVIDES_SM_72 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_72)) #define _NV_TARGET___NV_PROVIDES_SM_75 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_75)) #define _NV_TARGET___NV_PROVIDES_SM_80 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_80)) #define _NV_TARGET___NV_PROVIDES_SM_86 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_86)) #define _NV_TARGET___NV_PROVIDES_SM_87 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_87)) #define _NV_TARGET___NV_PROVIDES_SM_89 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_89)) #define _NV_TARGET___NV_PROVIDES_SM_90 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_90)) #define _NV_TARGET___NV_IS_EXACTLY_SM_35 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_35)) #define _NV_TARGET___NV_IS_EXACTLY_SM_37 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_37)) #define _NV_TARGET___NV_IS_EXACTLY_SM_50 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_50)) #define _NV_TARGET___NV_IS_EXACTLY_SM_52 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_52)) #define _NV_TARGET___NV_IS_EXACTLY_SM_53 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_53)) #define _NV_TARGET___NV_IS_EXACTLY_SM_60 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_60)) #define _NV_TARGET___NV_IS_EXACTLY_SM_61 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_61)) #define _NV_TARGET___NV_IS_EXACTLY_SM_62 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_62)) #define _NV_TARGET___NV_IS_EXACTLY_SM_70 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_70)) #define _NV_TARGET___NV_IS_EXACTLY_SM_72 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_72)) #define _NV_TARGET___NV_IS_EXACTLY_SM_75 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_75)) #define _NV_TARGET___NV_IS_EXACTLY_SM_80 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_80)) #define _NV_TARGET___NV_IS_EXACTLY_SM_86 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_86)) #define _NV_TARGET___NV_IS_EXACTLY_SM_87 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_87)) #define _NV_TARGET___NV_IS_EXACTLY_SM_89 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_89)) #define _NV_TARGET___NV_IS_EXACTLY_SM_90 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_90)) #define NV_PROVIDES_SM_35 __NV_PROVIDES_SM_35 #define NV_PROVIDES_SM_37 __NV_PROVIDES_SM_37 #define NV_PROVIDES_SM_50 __NV_PROVIDES_SM_50 #define NV_PROVIDES_SM_52 __NV_PROVIDES_SM_52 #define NV_PROVIDES_SM_53 __NV_PROVIDES_SM_53 #define NV_PROVIDES_SM_60 __NV_PROVIDES_SM_60 #define NV_PROVIDES_SM_61 __NV_PROVIDES_SM_61 #define NV_PROVIDES_SM_62 __NV_PROVIDES_SM_62 #define NV_PROVIDES_SM_70 __NV_PROVIDES_SM_70 #define NV_PROVIDES_SM_72 __NV_PROVIDES_SM_72 #define NV_PROVIDES_SM_75 __NV_PROVIDES_SM_75 #define NV_PROVIDES_SM_80 __NV_PROVIDES_SM_80 #define NV_PROVIDES_SM_86 __NV_PROVIDES_SM_86 #define NV_PROVIDES_SM_87 __NV_PROVIDES_SM_87 #define NV_PROVIDES_SM_89 __NV_PROVIDES_SM_89 #define NV_PROVIDES_SM_90 __NV_PROVIDES_SM_90 #define NV_IS_EXACTLY_SM_35 __NV_IS_EXACTLY_SM_35 #define NV_IS_EXACTLY_SM_37 __NV_IS_EXACTLY_SM_37 #define NV_IS_EXACTLY_SM_50 __NV_IS_EXACTLY_SM_50 #define NV_IS_EXACTLY_SM_52 __NV_IS_EXACTLY_SM_52 #define NV_IS_EXACTLY_SM_53 __NV_IS_EXACTLY_SM_53 #define NV_IS_EXACTLY_SM_60 __NV_IS_EXACTLY_SM_60 #define NV_IS_EXACTLY_SM_61 __NV_IS_EXACTLY_SM_61 #define NV_IS_EXACTLY_SM_62 __NV_IS_EXACTLY_SM_62 #define NV_IS_EXACTLY_SM_70 __NV_IS_EXACTLY_SM_70 #define NV_IS_EXACTLY_SM_72 __NV_IS_EXACTLY_SM_72 #define NV_IS_EXACTLY_SM_75 __NV_IS_EXACTLY_SM_75 #define NV_IS_EXACTLY_SM_80 __NV_IS_EXACTLY_SM_80 #define NV_IS_EXACTLY_SM_86 __NV_IS_EXACTLY_SM_86 #define NV_IS_EXACTLY_SM_87 __NV_IS_EXACTLY_SM_87 #define NV_IS_EXACTLY_SM_89 __NV_IS_EXACTLY_SM_87 #define NV_IS_EXACTLY_SM_90 __NV_IS_EXACTLY_SM_90 #define NV_IS_HOST __NV_IS_HOST #define NV_IS_DEVICE __NV_IS_DEVICE #define NV_ANY_TARGET __NV_ANY_TARGET #define NV_NO_TARGET __NV_NO_TARGET #if defined(_NV_COMPILER_NVCXX) && defined(_NVHPC_CUDA) # define _NV_ARCH_COND(q) (_NV_TARGET_##q) # define _NV_BLOCK_EXPAND(...) _NV_REMOVE_PAREN(__VA_ARGS__) # define _NV_TARGET_IF(cond, t, ...) \ (if target _NV_ARCH_COND(cond) { \ _NV_BLOCK_EXPAND(t) \ } else { _NV_BLOCK_EXPAND(__VA_ARGS__) }) #elif defined(_NV_COMPILER_NVCC) || defined (_NV_COMPILER_CLANG_CUDA) # if (_NV_TARGET___NV_IS_EXACTLY_SM_35) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_35 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_35 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_37) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_37 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_37 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_50) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_50 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_50 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_52) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_52 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_52 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_53) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_53 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_53 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_60) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_60 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_60 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_61) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_61 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_61 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_62) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_62 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_62 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_70) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_70 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_70 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_72) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_72 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_72 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_75) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_75 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_75 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_80) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_80 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_80 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_86) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_86 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_86 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_87) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_87 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_87 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_89) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_89 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_89 0 # endif # if (_NV_TARGET___NV_IS_EXACTLY_SM_90) # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_90 1 # else # define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_90 0 # endif # if (_NV_TARGET_IS_HOST) # define _NV_TARGET_BOOL___NV_IS_HOST 1 # define _NV_TARGET_BOOL___NV_IS_DEVICE 0 # else # define _NV_TARGET_BOOL___NV_IS_HOST 0 # define _NV_TARGET_BOOL___NV_IS_DEVICE 1 # endif # define _NV_TARGET_BOOL___NV_ANY_TARGET 1 # define _NV_TARGET_BOOL___NV_NO_TARGET 0 # if (_NV_TARGET___NV_PROVIDES_SM_35) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_35 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_35 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_37) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_37 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_37 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_50) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_50 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_50 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_52) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_52 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_52 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_53) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_53 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_53 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_60) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_60 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_60 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_61) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_61 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_61 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_62) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_62 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_62 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_70) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_70 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_70 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_72) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_72 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_72 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_75) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_75 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_75 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_80) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_80 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_80 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_86) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_86 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_86 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_87) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_87 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_87 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_89) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_89 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_89 0 # endif # if (_NV_TARGET___NV_PROVIDES_SM_90) # define _NV_TARGET_BOOL___NV_PROVIDES_SM_90 1 # else # define _NV_TARGET_BOOL___NV_PROVIDES_SM_90 0 # endif # define _NV_ARCH_COND_CAT1(cond) _NV_TARGET_BOOL_##cond # define _NV_ARCH_COND_CAT(cond) _NV_EVAL(_NV_ARCH_COND_CAT1(cond)) # define _NV_TARGET_EMPTY_PARAM ; # if defined(_NV_TARGET_CPP11) # define _NV_BLOCK_EXPAND(...) { _NV_REMOVE_PAREN(__VA_ARGS__) } # define _NV_TARGET_IF(cond, t, ...) _NV_IF( _NV_ARCH_COND_CAT(cond), t, __VA_ARGS__) # else # define _NV_BLOCK_EXPAND(x) { x } # define _NV_TARGET_IF(cond, t) _NV_IF(_NV_ARCH_COND_CAT(cond), t, _NV_TARGET_EMPTY_PARAM) # define _NV_TARGET_IF_ELSE(cond, t, f) _NV_IF(_NV_ARCH_COND_CAT(cond), t, f) # endif #endif #if defined(_NV_TARGET_CPP11) # define _NV_TARGET_DISPATCH_HANDLE0() # define _NV_TARGET_DISPATCH_HANDLE2(q, fn) _NV_TARGET_IF(q, fn) # define _NV_TARGET_DISPATCH_HANDLE4(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE2(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE6(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE4(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE8(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE6(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE10(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE8(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE12(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE10(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE14(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE12(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE16(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE14(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE18(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE16(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE20(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE18(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE22(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE20(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE24(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE22(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE26(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE24(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE28(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE26(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE30(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE28(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE32(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE30(__VA_ARGS__)) # define _NV_TARGET_DISPATCH_HANDLE34(q, fn, ...) _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE32(__VA_ARGS__)) # define _NV_TARGET_DISPATCH(...) _NV_BLOCK_EXPAND(_NV_DISPATCH_N_ARY(_NV_TARGET_DISPATCH_HANDLE, __VA_ARGS__)) # define NV_IF_TARGET(cond, ...) _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, __VA_ARGS__)) # define NV_IF_ELSE_TARGET(cond, t, f) _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, t, f)) # define NV_DISPATCH_TARGET(...) _NV_TARGET_DISPATCH(__VA_ARGS__) #else # define NV_IF_TARGET(cond, t) _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, t)) # define NV_IF_ELSE_TARGET(cond, t, f) _NV_BLOCK_EXPAND(_NV_TARGET_IF_ELSE(cond, t, f)) #endif #endif #endif #endif #if !defined(__VECTOR_FUNCTIONS_H__) #define __VECTOR_FUNCTIONS_H__ #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(__CUDACC_RTC__) #define __VECTOR_FUNCTIONS_DECL__ __host__ __device__ #else #define __VECTOR_FUNCTIONS_DECL__ static __inline__ __host__ __device__ #endif __VECTOR_FUNCTIONS_DECL__ char1 make_char1(signed char x); __VECTOR_FUNCTIONS_DECL__ uchar1 make_uchar1(unsigned char x); __VECTOR_FUNCTIONS_DECL__ char2 make_char2(signed char x, signed char y); __VECTOR_FUNCTIONS_DECL__ uchar2 make_uchar2(unsigned char x, unsigned char y); __VECTOR_FUNCTIONS_DECL__ char3 make_char3(signed char x, signed char y, signed char z); __VECTOR_FUNCTIONS_DECL__ uchar3 make_uchar3(unsigned char x, unsigned char y, unsigned char z); __VECTOR_FUNCTIONS_DECL__ char4 make_char4(signed char x, signed char y, signed char z, signed char w); __VECTOR_FUNCTIONS_DECL__ uchar4 make_uchar4(unsigned char x, unsigned char y, unsigned char z, unsigned char w); __VECTOR_FUNCTIONS_DECL__ short1 make_short1(short x); __VECTOR_FUNCTIONS_DECL__ ushort1 make_ushort1(unsigned short x); __VECTOR_FUNCTIONS_DECL__ short2 make_short2(short x, short y); __VECTOR_FUNCTIONS_DECL__ ushort2 make_ushort2(unsigned short x, unsigned short y); __VECTOR_FUNCTIONS_DECL__ short3 make_short3(short x,short y, short z); __VECTOR_FUNCTIONS_DECL__ ushort3 make_ushort3(unsigned short x, unsigned short y, unsigned short z); __VECTOR_FUNCTIONS_DECL__ short4 make_short4(short x, short y, short z, short w); __VECTOR_FUNCTIONS_DECL__ ushort4 make_ushort4(unsigned short x, unsigned short y, unsigned short z, unsigned short w); __VECTOR_FUNCTIONS_DECL__ int1 make_int1(int x); __VECTOR_FUNCTIONS_DECL__ uint1 make_uint1(unsigned int x); __VECTOR_FUNCTIONS_DECL__ int2 make_int2(int x, int y); __VECTOR_FUNCTIONS_DECL__ uint2 make_uint2(unsigned int x, unsigned int y); __VECTOR_FUNCTIONS_DECL__ int3 make_int3(int x, int y, int z); __VECTOR_FUNCTIONS_DECL__ uint3 make_uint3(unsigned int x, unsigned int y, unsigned int z); __VECTOR_FUNCTIONS_DECL__ int4 make_int4(int x, int y, int z, int w); __VECTOR_FUNCTIONS_DECL__ uint4 make_uint4(unsigned int x, unsigned int y, unsigned int z, unsigned int w); __VECTOR_FUNCTIONS_DECL__ long1 make_long1(long int x); __VECTOR_FUNCTIONS_DECL__ ulong1 make_ulong1(unsigned long int x); __VECTOR_FUNCTIONS_DECL__ long2 make_long2(long int x, long int y); __VECTOR_FUNCTIONS_DECL__ ulong2 make_ulong2(unsigned long int x, unsigned long int y); __VECTOR_FUNCTIONS_DECL__ long3 make_long3(long int x, long int y, long int z); __VECTOR_FUNCTIONS_DECL__ ulong3 make_ulong3(unsigned long int x, unsigned long int y, unsigned long int z); __VECTOR_FUNCTIONS_DECL__ long4 make_long4(long int x, long int y, long int z, long int w); __VECTOR_FUNCTIONS_DECL__ ulong4 make_ulong4(unsigned long int x, unsigned long int y, unsigned long int z, unsigned long int w); __VECTOR_FUNCTIONS_DECL__ float1 make_float1(float x); __VECTOR_FUNCTIONS_DECL__ float2 make_float2(float x, float y); __VECTOR_FUNCTIONS_DECL__ float3 make_float3(float x, float y, float z); __VECTOR_FUNCTIONS_DECL__ float4 make_float4(float x, float y, float z, float w); __VECTOR_FUNCTIONS_DECL__ longlong1 make_longlong1(long long int x); __VECTOR_FUNCTIONS_DECL__ ulonglong1 make_ulonglong1(unsigned long long int x); __VECTOR_FUNCTIONS_DECL__ longlong2 make_longlong2(long long int x, long long int y); __VECTOR_FUNCTIONS_DECL__ ulonglong2 make_ulonglong2(unsigned long long int x, unsigned long long int y); __VECTOR_FUNCTIONS_DECL__ longlong3 make_longlong3(long long int x, long long int y, long long int z); __VECTOR_FUNCTIONS_DECL__ ulonglong3 make_ulonglong3(unsigned long long int x, unsigned long long int y, unsigned long long int z); __VECTOR_FUNCTIONS_DECL__ longlong4 make_longlong4(long long int x, long long int y, long long int z, long long int w); __VECTOR_FUNCTIONS_DECL__ ulonglong4 make_ulonglong4(unsigned long long int x, unsigned long long int y, unsigned long long int z, unsigned long long int w); __VECTOR_FUNCTIONS_DECL__ double1 make_double1(double x); __VECTOR_FUNCTIONS_DECL__ double2 make_double2(double x, double y); __VECTOR_FUNCTIONS_DECL__ double3 make_double3(double x, double y, double z); __VECTOR_FUNCTIONS_DECL__ double4 make_double4(double x, double y, double z, double w); #undef __VECTOR_FUNCTIONS_DECL__ #if !defined(__CUDACC_RTC__) #include "vector_functions.hpp" #endif #endif #if defined(__CUDACC__) #if defined(__CUDACC_RTC__) #ifndef __NVRTC_DEVICE_RUNTIME_H__ #define __NVRTC_DEVICE_RUNTIME_H__ #if __cplusplus >= 201103L namespace __nvrtc_util { template<typename T> struct remove_ref { typedef T type; }; template<typename T> struct remove_ref<T&> { typedef T type; }; template<typename T> struct remove_ref<T&&> { typedef T type; }; template<class T> struct is_lvalue_ref { static const bool val = false; }; template<class T> struct is_lvalue_ref<T&> { static const bool val = true; }; } #if __NV_BUILTIN_MOVE_FORWARD namespace std { template<typename T> constexpr T&& __device__ forward(typename __nvrtc_util::remove_ref<T>::type &__in) noexcept { return static_cast<T&&>(__in); } template<typename T> constexpr T&& __device__ forward(typename __nvrtc_util::remove_ref<T>::type && __in) noexcept { static_assert(!__nvrtc_util::is_lvalue_ref<T>::val, "incorrect forward call!"); return static_cast<T&&>(__in); } template<typename T> constexpr typename __nvrtc_util::remove_ref<T>::type && __device__ move(T&& __in) noexcept { return static_cast<typename __nvrtc_util::remove_ref<T>::type &&>(__in); } } #endif #if __NV_BUILTIN_INITIALIZER_LIST namespace std { template<class T> class initializer_list { public: typedef T value_type; typedef const T& reference; typedef const T& const_reference; typedef size_t size_type; typedef const T* iterator; typedef const T* const_iterator; private: const T* buf_ptr; size_t length; constexpr initializer_list(const_iterator in, size_type in_length) : buf_ptr(in), length(in_length) { } public: constexpr initializer_list() noexcept : buf_ptr(0), length(0) { } constexpr size_type size() const noexcept { return length; } constexpr const_iterator begin() const noexcept { return buf_ptr; } constexpr const_iterator end() const noexcept { return buf_ptr + length; } }; } #endif #endif extern "C" inline __host__ __device__ void __cxa_vec_ctor(void *n, size_t num, size_t size, void (*c) (void *), void (*d) (void *)) { if (!c) return; for (size_t i = 0; i < num; i++) c((void*)((unsigned char *)n + i*size)); } extern "C" inline __host__ __device__ void __cxa_vec_cctor(void *dest, void *src, size_t num, size_t size, void (*c) (void *, void *), void (*d) (void *)) { if (!c) return; for (size_t i = 0; i < num; i++) { c((void*)((unsigned char *)dest + i*size), (void*)((unsigned char *)src + i*size)); } } extern "C" inline __host__ __device__ void __cxa_vec_dtor(void *n, size_t num, size_t size, void (*d) (void *)) { if (!d) return; for (size_t i = num-1; i > 0; i--) d((void*)((unsigned char *)n + i*size)); d((void*)((unsigned char *)n + 0*size)); } extern "C" inline __host__ __device__ void * __cxa_vec_new2(size_t num, size_t size, size_t pad, void (*c) (void *), void (*d) (void *), void *(*m) (size_t), void (*f) (void *)) { unsigned char *t = (unsigned char *)m(num*size + pad); *(size_t*)t = num; (void)__cxa_vec_ctor((void *)(t+pad), num, size, c, d); return (void *)(t+pad); } #if defined(_MSC_VER) extern "C" __host__ __device__ __cudart_builtin__ void* malloc(size_t); extern "C" __host__ __device__ __cudart_builtin__ void free(void*); #else extern "C" __host__ __device__ __cudart_builtin__ __attribute__((used)) void* malloc(size_t); extern "C" __host__ __device__ __cudart_builtin__ __attribute__((used)) void free(void*); #endif inline __host__ __device__ void *__nvrtc_malloc_wrapper(size_t in) { return malloc(in); } inline __host__ __device__ void __nvrtc_free_wrapper(void *in) { free(in); } extern "C" inline __host__ __device__ void * __cxa_vec_new(size_t num, size_t size, size_t pad, void (*c) (void *), void (*d) (void *)) { return __cxa_vec_new2(num, size, pad, c, d, __nvrtc_malloc_wrapper, __nvrtc_free_wrapper); } extern "C" inline __host__ __device__ void * __cxa_vec_new3(size_t num, size_t size, size_t pad, void (*c) (void *), void (*d) (void *), void *(*m) (size_t), void (*f)(void *, size_t)) { return __cxa_vec_new2(num, size, pad, c, d, m, 0); } extern "C" inline __host__ __device__ void __cxa_vec_delete2(void *n, size_t size, size_t pad, void (*d) (void *), void (*f) (void *)) { unsigned char *ptr = (unsigned char *)(n); if (ptr) { unsigned char *t = ptr - pad; size_t num = *(size_t*)t; __cxa_vec_dtor(ptr, num, size, d); f((void *)t); } } extern "C" inline __host__ __device__ void __cxa_vec_delete(void *n, size_t size, size_t pad, void (*d) (void *)) { __cxa_vec_delete2(n, size, pad, d, __nvrtc_free_wrapper); } extern "C" inline __host__ __device__ void __cxa_vec_delete3(void *n, size_t size, size_t pad, void (*d)(void *), void (*f)(void *,size_t)) { unsigned char *ptr = (unsigned char *)(n); if (ptr) { unsigned char *t = ptr - pad; size_t num = *(size_t*)t; size_t tsize = num*size+pad; __cxa_vec_dtor(ptr, num, size, d); f((void *)t, tsize); } } extern "C" inline __host__ __device__ void *__gen_nvvm_memcpy_aligned1(void *dest, void *src, size_t size) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 1); return dest; } extern "C" inline __host__ __device__ void *__gen_nvvm_memcpy_aligned2(void *dest, void *src, size_t size) { if (((unsigned long long)(dest) % 2) == 0 && ((unsigned long long)(src) % 2) == 0) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 2); } else { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 1); } return dest; } extern "C" inline __host__ __device__ void *__gen_nvvm_memcpy_aligned4(void *dest, void *src, size_t size) { if (((unsigned long long)(dest) % 4) == 0 && ((unsigned long long)(src) % 4) == 0) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 4); } else { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 1); } return dest; } extern "C" inline __host__ __device__ void * __gen_nvvm_memcpy_aligned8(void *dest, void *src, size_t size) { if (((unsigned long long)(dest) % 8) == 0 && ((unsigned long long)(src) % 8) == 0) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 8); } else { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 1); } return dest; } extern "C" inline __host__ __device__ void * __gen_nvvm_memcpy_aligned16(void *dest, void *src, size_t size) { if (((unsigned long long)(dest) % 16) == 0 && ((unsigned long long)(src) % 16) == 0) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 16); } else { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, size, 1); } return dest; } extern "C" inline __host__ __device__ void __cxa_pure_virtual(void) { volatile int *ptr = 0; *ptr = 1; } extern "C" inline __host__ __device__ void* __gen_nvvm_memset(void *dest, int c, size_t n) { __nvvm_memset((unsigned char *)dest, (unsigned char)c, n, 1); return dest; } extern "C" inline __host__ __device__ void* __gen_nvvm_memcpy(void *dest, const void *src, size_t n) { __nvvm_memcpy((unsigned char *)dest, (unsigned char *)src, n, 1); return dest; } #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/device_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/device_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_FUNCTIONS_H__ #endif #if !defined(__DEVICE_FUNCTIONS_H__) #define __DEVICE_FUNCTIONS_H__ #if defined(__cplusplus) && defined(__CUDACC__) #if defined(__CUDACC_RTC__) #define __DEVICE_FUNCTIONS_DECL__ __device__ __cudart_builtin__ #define __DEVICE_FUNCTIONS_STATIC_DECL__ __device__ __cudart_builtin__ #define __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ __device__ __host__ __cudart_builtin__ #else #define __DEVICE_FUNCTIONS_DECL__ __device__ __cudart_builtin__ #define __DEVICE_FUNCTIONS_STATIC_DECL__ static __inline__ __device__ __cudart_builtin__ #define __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ static __inline__ __device__ __host__ __cudart_builtin__ #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif extern "C" { __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __mulhi(int x, int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __umulhi(unsigned int x, unsigned int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __mul64hi(long long int x, long long int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __umul64hi(unsigned long long int x, unsigned long long int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __int_as_float(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __float_as_int(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __uint_as_float(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __float_as_uint(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void __syncthreads(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void __prof_trigger(int); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void __threadfence(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void __threadfence_block(void); __DEVICE_FUNCTIONS_DECL__ #if defined(__GNUC__) || defined(__CUDACC_RTC__) __attribute__((__noreturn__)) #elif defined(_MSC_VER) __declspec(noreturn) #endif __device_builtin__ void __trap(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void __brkpt(); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __saturatef(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __sad(int x, int y, unsigned int z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __usad(unsigned int x, unsigned int y, unsigned int z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __mul24(int x, int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __umul24(unsigned int x, unsigned int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fdividef(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fdividef(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double fdivide(double x, double y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __sinf(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __cosf(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __tanf(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ void __sincosf(float x, float *sptr, float *cptr) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __expf(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __exp10f(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __log2f(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __log10f(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __logf(float x) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ float __powf(float x, float y) __THROW; __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __float2int_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __float2int_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __float2int_ru(float); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __float2int_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __float2uint_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __float2uint_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __float2uint_ru(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __float2uint_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __int2float_rn(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __int2float_rz(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __int2float_ru(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __int2float_rd(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __uint2float_rn(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __uint2float_rz(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __uint2float_ru(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __uint2float_rd(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __float2ll_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __float2ll_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __float2ll_ru(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __float2ll_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __float2ull_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __float2ull_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __float2ull_ru(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __float2ull_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ll2float_rn(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ll2float_rz(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ll2float_ru(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ll2float_rd(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ull2float_rn(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ull2float_rz(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ull2float_ru(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __ull2float_rd(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fadd_rn(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fadd_rz(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fadd_ru(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fadd_rd(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsub_rn(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsub_rz(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsub_ru(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsub_rd(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmul_rn(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmul_rz(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmul_ru(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmul_rd(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmaf_rn(float x, float y, float z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmaf_rz(float x, float y, float z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmaf_ru(float x, float y, float z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fmaf_rd(float x, float y, float z); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __frcp_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __frcp_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __frcp_ru(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __frcp_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsqrt_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsqrt_rz(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsqrt_ru(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fsqrt_rd(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __frsqrt_rn(float x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fdiv_rn(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fdiv_rz(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fdiv_ru(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __fdiv_rd(float x, float y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __clz(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __ffs(int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __popc(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __brev(unsigned int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __clzll(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __ffsll(long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __popcll(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __brevll(unsigned long long int x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __byte_perm(unsigned int x, unsigned int y, unsigned int s); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __hadd(int x, int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __rhadd(int x, int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __uhadd(unsigned int x, unsigned int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __urhadd(unsigned int x, unsigned int y); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __double2int_rz(double x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __double2uint_rz(double x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int __double2ll_rz(double x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int __double2ull_rz(double x); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __pm0(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __pm1(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __pm2(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __pm3(void); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabs2(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsss2(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vadd2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vaddss2 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vaddus2 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vavgs2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vavgu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vhaddu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpeq2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpges2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgeu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgts2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgtu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmples2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpleu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmplts2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpltu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpne2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsdiffu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmaxs2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmaxu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmins2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vminu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vseteq2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetges2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgeu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgts2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgtu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetles2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetleu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetlts2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetltu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetne2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsadu2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsub2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsubss2 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsubus2 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vneg2(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vnegss2(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsdiffs2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsads2(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabs4(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsss4(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vadd4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vaddss4 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vaddus4 (unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vavgs4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vavgu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vhaddu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpeq4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpges4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgeu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgts4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpgtu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmples4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpleu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmplts4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpltu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vcmpne4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsdiffu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmaxs4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmaxu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vmins4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vminu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vseteq4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetles4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetleu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetlts4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetltu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetges4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgeu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgts4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetgtu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsetne4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsadu4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsub4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsubss4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsubus4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vneg4(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vnegss4(unsigned int a); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vabsdiffs4(unsigned int a, unsigned int b); __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int __vsads4(unsigned int a, unsigned int b); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimax_s32_relu(const int a, const int b); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimax_s16x2_relu(const unsigned int a, const unsigned int b); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimin_s32_relu(const int a, const int b); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimin_s16x2_relu(const unsigned int a, const unsigned int b); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimax3_s32(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimax3_s16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimax3_u32(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimax3_u16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimin3_s32(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimin3_s16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimin3_u32(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimin3_u16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimax3_s32_relu(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimax3_s16x2_relu(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vimin3_s32_relu(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vimin3_s16x2_relu(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __viaddmax_s32(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmax_s16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmax_u32(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmax_u16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __viaddmin_s32(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmin_s16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmin_u32(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmin_u16x2(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __viaddmax_s32_relu(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmax_s16x2_relu(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __viaddmin_s32_relu(const int a, const int b, const int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __viaddmin_s16x2_relu(const unsigned int a, const unsigned int b, const unsigned int c); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vibmax_s32(const int a, const int b, bool* const pred); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmax_u32(const unsigned int a, const unsigned int b, bool* const pred); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ int __vibmin_s32(const int a, const int b, bool* const pred); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmin_u32(const unsigned int a, const unsigned int b, bool* const pred); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmax_s16x2(const unsigned int a, const unsigned int b, bool* const pred_hi, bool* const pred_lo); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmax_u16x2(const unsigned int a, const unsigned int b, bool* const pred_hi, bool* const pred_lo); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmin_s16x2(const unsigned int a, const unsigned int b, bool* const pred_hi, bool* const pred_lo); __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ unsigned int __vibmin_u16x2(const unsigned int a, const unsigned int b, bool* const pred_hi, bool* const pred_lo); } #undef __DEVICE_FUNCTIONS_DECL__ #undef __DEVICE_FUNCTIONS_STATIC_DECL__ #undef __DEVICE_HOST_FUNCTIONS_STATIC_DECL__ #endif #if !defined(__CUDACC_RTC__) #include "device_functions.hpp" #endif #if !defined(__DEVICE_ATOMIC_FUNCTIONS_H__) #define __DEVICE_ATOMIC_FUNCTIONS_H__ #if defined(__CUDACC_RTC__) #define __DEVICE_ATOMIC_FUNCTIONS_DECL__ __device__ #elif defined(_NVHPC_CUDA) # define __DEVICE_ATOMIC_FUNCTIONS_DECL__ extern __device__ __cudart_builtin__ #else #define __DEVICE_ATOMIC_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if !defined(__CUDA_ARCH__) && !defined(_NVHPC_CUDA) #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA) extern "C" { extern __device__ __device_builtin__ int __iAtomicAdd(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicAdd(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicExch(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicExch(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ float __fAtomicExch(float *address, float val); extern __device__ __device_builtin__ int __iAtomicMin(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicMin(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicMax(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicMax(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicInc(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicDec(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicAnd(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicAnd(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicOr(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicOr(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicXor(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicXor(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicCAS(int *address, int compare, int val); extern __device__ __device_builtin__ unsigned int __uAtomicCAS(unsigned int *address, unsigned int compare, unsigned int val); } #endif __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicAdd(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAdd(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicSub(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicSub(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicExch(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicExch(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ float atomicExch(float *address, float val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicMin(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMin(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicMax(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMax(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicInc(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicDec(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicAnd(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAnd(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicOr(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicOr(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicXor(int *address, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicXor(unsigned int *address, unsigned int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ int atomicCAS(int *address, int compare, int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicCAS(unsigned int *address, unsigned int compare, unsigned int val) __DEF_IF_HOST #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 700 #define __WSB_DEPRECATION_MESSAGE(x) #x"() is not valid on compute_70 and above, and should be replaced with "#x"_sync()."\ "To continue using "#x"(), specify virtual architecture compute_60 when targeting sm_70 and above, for example, using the pair of compiler options: -arch=compute_60 -code=sm_70." #elif defined(_NVHPC_CUDA) #define __WSB_DEPRECATION_MESSAGE(x) #x"() is not valid on cc70 and above, and should be replaced with "#x"_sync()." #else #define __WSB_DEPRECATION_MESSAGE(x) #x"() is deprecated in favor of "#x"_sync() and may be removed in a future release (Use -Wno-deprecated-declarations to suppress this warning)." #endif extern "C" { #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA) extern __device__ __device_builtin__ unsigned long long int __ullAtomicAdd(unsigned long long int *address, unsigned long long int val); extern __device__ __device_builtin__ unsigned long long int __ullAtomicExch(unsigned long long int *address, unsigned long long int val); extern __device__ __device_builtin__ unsigned long long int __ullAtomicCAS(unsigned long long int *address, unsigned long long int compare, unsigned long long int val); #endif extern __device__ __device_builtin__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__any)) int __any(int cond); extern __device__ __device_builtin__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__all)) int __all(int cond); } __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned long long int atomicAdd(unsigned long long int *address, unsigned long long int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned long long int atomicExch(unsigned long long int *address, unsigned long long int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ unsigned long long int atomicCAS(unsigned long long int *address, unsigned long long int compare, unsigned long long int val) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__any)) bool any(bool cond) __DEF_IF_HOST __DEVICE_ATOMIC_FUNCTIONS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__all)) bool all(bool cond) __DEF_IF_HOST #undef __DEPRECATED__ #undef __WSB_DEPRECATION_MESSAGE #endif #undef __DEF_IF_HOST #undef __DEVICE_ATOMIC_FUNCTIONS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "device_atomic_functions.hpp" #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/device_double_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/device_double_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_DOUBLE_FUNCTIONS_H__ #endif #if !defined(__DEVICE_DOUBLE_FUNCTIONS_H__) #define __DEVICE_DOUBLE_FUNCTIONS_H__ #if defined(__cplusplus) && defined(__CUDACC__) #if defined(__CUDACC_RTC__) #define __DEVICE_DOUBLE_FUNCTIONS_DECL__ __device__ #else #define __DEVICE_DOUBLE_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif extern "C" { extern __device__ __device_builtin__ long long int __double_as_longlong(double x); extern __device__ __device_builtin__ double __longlong_as_double(long long int x); extern __device__ __device_builtin__ double __fma_rn(double x, double y, double z); extern __device__ __device_builtin__ double __fma_rz(double x, double y, double z); extern __device__ __device_builtin__ double __fma_ru(double x, double y, double z); extern __device__ __device_builtin__ double __fma_rd(double x, double y, double z); extern __device__ __device_builtin__ double __dadd_rn(double x, double y); extern __device__ __device_builtin__ double __dadd_rz(double x, double y); extern __device__ __device_builtin__ double __dadd_ru(double x, double y); extern __device__ __device_builtin__ double __dadd_rd(double x, double y); extern __device__ __device_builtin__ double __dsub_rn(double x, double y); extern __device__ __device_builtin__ double __dsub_rz(double x, double y); extern __device__ __device_builtin__ double __dsub_ru(double x, double y); extern __device__ __device_builtin__ double __dsub_rd(double x, double y); extern __device__ __device_builtin__ double __dmul_rn(double x, double y); extern __device__ __device_builtin__ double __dmul_rz(double x, double y); extern __device__ __device_builtin__ double __dmul_ru(double x, double y); extern __device__ __device_builtin__ double __dmul_rd(double x, double y); extern __device__ __device_builtin__ float __double2float_rn(double x); extern __device__ __device_builtin__ float __double2float_rz(double x); extern __device__ __device_builtin__ float __double2float_ru(double x); extern __device__ __device_builtin__ float __double2float_rd(double x); extern __device__ __device_builtin__ int __double2int_rn(double x); extern __device__ __device_builtin__ int __double2int_ru(double x); extern __device__ __device_builtin__ int __double2int_rd(double x); extern __device__ __device_builtin__ unsigned int __double2uint_rn(double x); extern __device__ __device_builtin__ unsigned int __double2uint_ru(double x); extern __device__ __device_builtin__ unsigned int __double2uint_rd(double x); extern __device__ __device_builtin__ long long int __double2ll_rn(double x); extern __device__ __device_builtin__ long long int __double2ll_ru(double x); extern __device__ __device_builtin__ long long int __double2ll_rd(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_rn(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_ru(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_rd(double x); extern __device__ __device_builtin__ double __int2double_rn(int x); extern __device__ __device_builtin__ double __uint2double_rn(unsigned int x); extern __device__ __device_builtin__ double __ll2double_rn(long long int x); extern __device__ __device_builtin__ double __ll2double_rz(long long int x); extern __device__ __device_builtin__ double __ll2double_ru(long long int x); extern __device__ __device_builtin__ double __ll2double_rd(long long int x); extern __device__ __device_builtin__ double __ull2double_rn(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_rz(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_ru(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_rd(unsigned long long int x); extern __device__ __device_builtin__ int __double2hiint(double x); extern __device__ __device_builtin__ int __double2loint(double x); extern __device__ __device_builtin__ double __hiloint2double(int hi, int lo); } __DEVICE_DOUBLE_FUNCTIONS_DECL__ double fma(double a, double b, double c, enum cudaRoundMode mode); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double dmul(double a, double b, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double dadd(double a, double b, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double dsub(double a, double b, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ int double2int(double a, enum cudaRoundMode mode = cudaRoundZero); __DEVICE_DOUBLE_FUNCTIONS_DECL__ unsigned int double2uint(double a, enum cudaRoundMode mode = cudaRoundZero); __DEVICE_DOUBLE_FUNCTIONS_DECL__ long long int double2ll(double a, enum cudaRoundMode mode = cudaRoundZero); __DEVICE_DOUBLE_FUNCTIONS_DECL__ unsigned long long int double2ull(double a, enum cudaRoundMode mode = cudaRoundZero); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double ll2double(long long int a, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double ull2double(unsigned long long int a, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double int2double(int a, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double uint2double(unsigned int a, enum cudaRoundMode mode = cudaRoundNearest); __DEVICE_DOUBLE_FUNCTIONS_DECL__ double float2double(float a, enum cudaRoundMode mode = cudaRoundNearest); #undef __DEVICE_DOUBLE_FUNCTIONS_DECL__ #endif #if !defined(__CUDACC_RTC__) #include "device_double_functions.hpp" #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_DOUBLE_FUNCTIONS_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_DOUBLE_FUNCTIONS_H__ #endif #if !defined(__SM_20_ATOMIC_FUNCTIONS_H__) #define __SM_20_ATOMIC_FUNCTIONS_H__ #if defined(__CUDACC_RTC__) #define __SM_20_ATOMIC_FUNCTIONS_DECL__ __device__ #elif defined(_NVHPC_CUDA) #define __SM_20_ATOMIC_FUNCTIONS_DECL__ extern __device__ __cudart_builtin__ #else #define __SM_20_ATOMIC_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ __location__(device) __location__(host) #endif #if !defined(__CUDA_ARCH__) && !defined(_NVHPC_CUDA) #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA) extern "C" { extern __device__ __device_builtin__ float __fAtomicAdd(float *address, float val); } #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ #endif __SM_20_ATOMIC_FUNCTIONS_DECL__ float atomicAdd(float *address, float val) __DEF_IF_HOST #endif #undef __DEF_IF_HOST #undef __SM_20_ATOMIC_FUNCTIONS_DECL__ #if !defined(__CUDACC_RTC__) && (defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA)) #include "sm_20_atomic_functions.hpp" #endif #endif #if !defined(__SM_32_ATOMIC_FUNCTIONS_H__) #define __SM_32_ATOMIC_FUNCTIONS_H__ #if defined(__CUDACC_RTC__) #define __SM_32_ATOMIC_FUNCTIONS_DECL__ __device__ #else #define __SM_32_ATOMIC_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ __location__(device) __location__(host) #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #ifdef __CUDA_ARCH__ extern "C" { extern __device__ __device_builtin__ long long __illAtomicMin(long long *address, long long val); extern __device__ __device_builtin__ long long __illAtomicMax(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicAnd(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicOr(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicXor(long long *address, long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMin(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMax(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAnd(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicOr (unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicXor(unsigned long long *address, unsigned long long val); } #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ #endif __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicMin(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicMax(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicAnd(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicOr(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicXor(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMin(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMax(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAnd(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicOr(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicXor(unsigned long long *address, unsigned long long val) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_32_ATOMIC_FUNCTIONS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_32_atomic_functions.hpp" #endif #endif #if !defined(__SM_35_ATOMIC_FUNCTIONS_H__) #define __SM_35_ATOMIC_FUNCTIONS_H__ #if !defined(__SM_32_ATOMIC_FUNCTIONS_H__) #define __SM_32_ATOMIC_FUNCTIONS_H__ #if defined(__CUDACC_RTC__) #define __SM_32_ATOMIC_FUNCTIONS_DECL__ __device__ #else #define __SM_32_ATOMIC_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ __location__(device) __location__(host) #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #ifdef __CUDA_ARCH__ extern "C" { extern __device__ __device_builtin__ long long __illAtomicMin(long long *address, long long val); extern __device__ __device_builtin__ long long __illAtomicMax(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicAnd(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicOr(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicXor(long long *address, long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMin(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMax(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAnd(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicOr (unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicXor(unsigned long long *address, unsigned long long val); } #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ #endif __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicMin(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicMax(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicAnd(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicOr(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ long long atomicXor(long long *address, long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMin(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMax(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAnd(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicOr(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_32_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicXor(unsigned long long *address, unsigned long long val) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_32_ATOMIC_FUNCTIONS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_32_atomic_functions.hpp" #endif #endif #endif #if !defined(__SM_60_ATOMIC_FUNCTIONS_H__) #define __SM_60_ATOMIC_FUNCTIONS_H__ #if defined(__CUDACC_RTC__) #define __SM_60_ATOMIC_FUNCTIONS_DECL__ __device__ #elif defined(_NVHPC_CUDA) #define __SM_60_ATOMIC_FUNCTIONS_DECL__ extern __device__ __cudart_builtin__ #else #define __SM_60_ATOMIC_FUNCTIONS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 600 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if !defined(__CUDA_ARCH__) && !defined(_NVHPC_CUDA) #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #ifdef __CUDA_ARCH__ extern "C" { extern __device__ __device_builtin__ double __dAtomicAdd(double *address, double val); extern __device__ __device_builtin__ int __iAtomicAdd_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicAdd_system(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicAdd_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicAdd_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAdd_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAdd_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ float __fAtomicAdd_block(float *address, float val); extern __device__ __device_builtin__ float __fAtomicAdd_system(float *address, float val); extern __device__ __device_builtin__ double __dAtomicAdd_block(double *address, double val); extern __device__ __device_builtin__ double __dAtomicAdd_system(double *address, double val); extern __device__ __device_builtin__ int __iAtomicExch_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicExch_system(int *address, int val); extern __device__ __device_builtin__ unsigned int __uAtomicExch_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicExch_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicExch_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicExch_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ float __fAtomicExch_block(float *address, float val); extern __device__ __device_builtin__ float __fAtomicExch_system(float *address, float val); extern __device__ __device_builtin__ int __iAtomicMin_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicMin_system(int *address, int val); extern __device__ __device_builtin__ long long __illAtomicMin_block(long long *address, long long val); extern __device__ __device_builtin__ long long __illAtomicMin_system(long long *address, long long val); extern __device__ __device_builtin__ unsigned int __uAtomicMin_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicMin_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMin_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMin_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ int __iAtomicMax_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicMax_system(int *address, int val); extern __device__ __device_builtin__ long long __illAtomicMax_block(long long *address, long long val); extern __device__ __device_builtin__ long long __illAtomicMax_system(long long *address, long long val); extern __device__ __device_builtin__ unsigned int __uAtomicMax_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicMax_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMax_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicMax_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned int __uAtomicInc_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicInc_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicDec_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicDec_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ int __iAtomicCAS_block(int *address, int compare, int val); extern __device__ __device_builtin__ int __iAtomicCAS_system(int *address, int compare, int val); extern __device__ __device_builtin__ unsigned int __uAtomicCAS_block(unsigned int *address, unsigned int compare, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicCAS_system(unsigned int *address, unsigned int compare, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicCAS_block(unsigned long long int *address, unsigned long long int compare, unsigned long long int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicCAS_system(unsigned long long int *address, unsigned long long int compare, unsigned long long int val); extern __device__ __device_builtin__ int __iAtomicAnd_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicAnd_system(int *address, int val); extern __device__ __device_builtin__ long long __llAtomicAnd_block(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicAnd_system(long long *address, long long val); extern __device__ __device_builtin__ unsigned int __uAtomicAnd_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicAnd_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAnd_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicAnd_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ int __iAtomicOr_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicOr_system(int *address, int val); extern __device__ __device_builtin__ long long __llAtomicOr_block(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicOr_system(long long *address, long long val); extern __device__ __device_builtin__ unsigned int __uAtomicOr_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicOr_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicOr_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicOr_system(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ int __iAtomicXor_block(int *address, int val); extern __device__ __device_builtin__ int __iAtomicXor_system(int *address, int val); extern __device__ __device_builtin__ long long __llAtomicXor_block(long long *address, long long val); extern __device__ __device_builtin__ long long __llAtomicXor_system(long long *address, long long val); extern __device__ __device_builtin__ unsigned int __uAtomicXor_block(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned int __uAtomicXor_system(unsigned int *address, unsigned int val); extern __device__ __device_builtin__ unsigned long long __ullAtomicXor_block(unsigned long long *address, unsigned long long val); extern __device__ __device_builtin__ unsigned long long __ullAtomicXor_system(unsigned long long *address, unsigned long long val); } #endif __SM_60_ATOMIC_FUNCTIONS_DECL__ double atomicAdd(double *address, double val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicAdd_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicAdd_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAdd_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAdd_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAdd_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAdd_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ float atomicAdd_block(float *address, float val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ float atomicAdd_system(float *address, float val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ double atomicAdd_block(double *address, double val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ double atomicAdd_system(double *address, double val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicSub_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicSub_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicSub_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicSub_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicExch_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicExch_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicExch_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicExch_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicExch_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicExch_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ float atomicExch_block(float *address, float val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ float atomicExch_system(float *address, float val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicMin_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicMin_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicMin_block(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicMin_system(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMin_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMin_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMin_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMin_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicMax_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicMax_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicMax_block(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicMax_system(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMax_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicMax_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMax_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicMax_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicInc_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicInc_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicDec_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicDec_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicCAS_block(int *address, int compare, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicCAS_system(int *address, int compare, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicCAS_block(unsigned int *address, unsigned int compare, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicCAS_system(unsigned int *address, unsigned int compare, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long int atomicCAS_block(unsigned long long int *address, unsigned long long int compare, unsigned long long int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long int atomicCAS_system(unsigned long long int *address, unsigned long long int compare, unsigned long long int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicAnd_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicAnd_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicAnd_block(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicAnd_system(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAnd_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicAnd_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAnd_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicAnd_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicOr_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicOr_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicOr_block(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicOr_system(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicOr_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicOr_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicOr_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicOr_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicXor_block(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ int atomicXor_system(int *address, int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicXor_block(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ long long atomicXor_system(long long *address, long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicXor_block(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned int atomicXor_system(unsigned int *address, unsigned int val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicXor_block(unsigned long long *address, unsigned long long val) __DEF_IF_HOST __SM_60_ATOMIC_FUNCTIONS_DECL__ unsigned long long atomicXor_system(unsigned long long *address, unsigned long long val) __DEF_IF_HOST #endif #endif #undef __SM_60_ATOMIC_FUNCTIONS_DECL__ #undef __DEF_IF_HOST #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_60_atomic_functions.hpp" #endif #endif #if !defined(__SM_20_INTRINSICS_H__) #define __SM_20_INTRINSICS_H__ #if defined(__CUDACC_RTC__) #define __SM_20_INTRINSICS_DECL__ __device__ #else #define __SM_20_INTRINSICS_DECL__ static __inline__ __device__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ __location__(device) __location__(host) #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #if defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= 700 #define __WSB_DEPRECATION_MESSAGE(x) #x"() is not valid on compute_70 and above, and should be replaced with "#x"_sync()."\ "To continue using "#x"(), specify virtual architecture compute_60 when targeting sm_70 and above, for example, using the pair of compiler options: -arch=compute_60 -code=sm_70." #elif defined(_NVHPC_CUDA) #define __WSB_DEPRECATION_MESSAGE(x) #x"() is not valid on cc70 and above, and should be replaced with "#x"_sync()." #else #define __WSB_DEPRECATION_MESSAGE(x) #x"() is deprecated in favor of "#x"_sync() and may be removed in a future release (Use -Wno-deprecated-declarations to suppress this warning)." #endif extern "C" { extern __device__ __device_builtin__ void __threadfence_system(void); extern __device__ __device_builtin__ double __ddiv_rn(double x, double y); extern __device__ __device_builtin__ double __ddiv_rz(double x, double y); extern __device__ __device_builtin__ double __ddiv_ru(double x, double y); extern __device__ __device_builtin__ double __ddiv_rd(double x, double y); extern __device__ __device_builtin__ double __drcp_rn(double x); extern __device__ __device_builtin__ double __drcp_rz(double x); extern __device__ __device_builtin__ double __drcp_ru(double x); extern __device__ __device_builtin__ double __drcp_rd(double x); extern __device__ __device_builtin__ double __dsqrt_rn(double x); extern __device__ __device_builtin__ double __dsqrt_rz(double x); extern __device__ __device_builtin__ double __dsqrt_ru(double x); extern __device__ __device_builtin__ double __dsqrt_rd(double x); extern __device__ __device_builtin__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__ballot)) unsigned int __ballot(int); extern __device__ __device_builtin__ int __syncthreads_count(int); extern __device__ __device_builtin__ int __syncthreads_and(int); extern __device__ __device_builtin__ int __syncthreads_or(int); extern __device__ __device_builtin__ long long int clock64(void); extern __device__ __device_builtin__ float __fmaf_ieee_rn(float x, float y, float z); extern __device__ __device_builtin__ float __fmaf_ieee_rd(float x, float y, float z); extern __device__ __device_builtin__ float __fmaf_ieee_ru(float x, float y, float z); extern __device__ __device_builtin__ float __fmaf_ieee_rz(float x, float y, float z); extern __device__ __device_builtin__ long long int __double_as_longlong(double x); extern __device__ __device_builtin__ double __longlong_as_double(long long int x); extern __device__ __device_builtin__ double __fma_rn(double x, double y, double z); extern __device__ __device_builtin__ double __fma_rz(double x, double y, double z); extern __device__ __device_builtin__ double __fma_ru(double x, double y, double z); extern __device__ __device_builtin__ double __fma_rd(double x, double y, double z); extern __device__ __device_builtin__ double __dadd_rn(double x, double y); extern __device__ __device_builtin__ double __dadd_rz(double x, double y); extern __device__ __device_builtin__ double __dadd_ru(double x, double y); extern __device__ __device_builtin__ double __dadd_rd(double x, double y); extern __device__ __device_builtin__ double __dsub_rn(double x, double y); extern __device__ __device_builtin__ double __dsub_rz(double x, double y); extern __device__ __device_builtin__ double __dsub_ru(double x, double y); extern __device__ __device_builtin__ double __dsub_rd(double x, double y); extern __device__ __device_builtin__ double __dmul_rn(double x, double y); extern __device__ __device_builtin__ double __dmul_rz(double x, double y); extern __device__ __device_builtin__ double __dmul_ru(double x, double y); extern __device__ __device_builtin__ double __dmul_rd(double x, double y); extern __device__ __device_builtin__ float __double2float_rn(double x); extern __device__ __device_builtin__ float __double2float_rz(double x); extern __device__ __device_builtin__ float __double2float_ru(double x); extern __device__ __device_builtin__ float __double2float_rd(double x); extern __device__ __device_builtin__ int __double2int_rn(double x); extern __device__ __device_builtin__ int __double2int_ru(double x); extern __device__ __device_builtin__ int __double2int_rd(double x); extern __device__ __device_builtin__ unsigned int __double2uint_rn(double x); extern __device__ __device_builtin__ unsigned int __double2uint_ru(double x); extern __device__ __device_builtin__ unsigned int __double2uint_rd(double x); extern __device__ __device_builtin__ long long int __double2ll_rn(double x); extern __device__ __device_builtin__ long long int __double2ll_ru(double x); extern __device__ __device_builtin__ long long int __double2ll_rd(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_rn(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_ru(double x); extern __device__ __device_builtin__ unsigned long long int __double2ull_rd(double x); extern __device__ __device_builtin__ double __int2double_rn(int x); extern __device__ __device_builtin__ double __uint2double_rn(unsigned int x); extern __device__ __device_builtin__ double __ll2double_rn(long long int x); extern __device__ __device_builtin__ double __ll2double_rz(long long int x); extern __device__ __device_builtin__ double __ll2double_ru(long long int x); extern __device__ __device_builtin__ double __ll2double_rd(long long int x); extern __device__ __device_builtin__ double __ull2double_rn(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_rz(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_ru(unsigned long long int x); extern __device__ __device_builtin__ double __ull2double_rd(unsigned long long int x); extern __device__ __device_builtin__ int __double2hiint(double x); extern __device__ __device_builtin__ int __double2loint(double x); extern __device__ __device_builtin__ double __hiloint2double(int hi, int lo); } #if defined(_NVHPC_CUDA) #undef __device_builtin__ #define __device_builtin__ #endif __SM_20_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__ballot)) unsigned int ballot(bool pred) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ int syncthreads_count(bool pred) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ bool syncthreads_and(bool pred) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ bool syncthreads_or(bool pred) __DEF_IF_HOST #undef __DEPRECATED__ #undef __WSB_DEPRECATION_MESSAGE __SM_20_INTRINSICS_DECL__ unsigned int __isGlobal(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ unsigned int __isShared(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ unsigned int __isConstant(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ unsigned int __isLocal(const void *ptr) __DEF_IF_HOST #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 700) __SM_20_INTRINSICS_DECL__ unsigned int __isGridConstant(const void *ptr) __DEF_IF_HOST #endif __SM_20_INTRINSICS_DECL__ size_t __cvta_generic_to_global(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ size_t __cvta_generic_to_shared(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ size_t __cvta_generic_to_constant(const void *ptr) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ size_t __cvta_generic_to_local(const void *ptr) __DEF_IF_HOST #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 700) __SM_20_INTRINSICS_DECL__ size_t __cvta_generic_to_grid_constant(const void *ptr) __DEF_IF_HOST #endif __SM_20_INTRINSICS_DECL__ void * __cvta_global_to_generic(size_t rawbits) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ void * __cvta_shared_to_generic(size_t rawbits) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ void * __cvta_constant_to_generic(size_t rawbits) __DEF_IF_HOST __SM_20_INTRINSICS_DECL__ void * __cvta_local_to_generic(size_t rawbits) __DEF_IF_HOST #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 700) __SM_20_INTRINSICS_DECL__ void * __cvta_grid_constant_to_generic(size_t rawbits) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_20_INTRINSICS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_20_intrinsics.hpp" #endif #endif #if !defined(__SM_30_INTRINSICS_H__) #define __SM_30_INTRINSICS_H__ #if defined(__CUDACC_RTC__) #define __SM_30_INTRINSICS_DECL__ __device__ #elif defined(_NVHPC_CUDA) #define __SM_30_INTRINSICS_DECL__ extern __device__ __cudart_builtin__ #else #define __SM_30_INTRINSICS_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 300 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if !defined(__CUDA_ARCH__) && !defined(_NVHPC_CUDA) #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif #if !defined warpSize && !defined __local_warpSize #define warpSize 32 #define __local_warpSize #endif #if defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 700 #define __WSB_DEPRECATION_MESSAGE(x) #x"() is deprecated in favor of "#x"_sync() and may be removed in a future release (Use -Wno-deprecated-declarations to suppress this warning)." #elif defined(_NVHPC_CUDA) #define __WSB_DEPRECATION_MESSAGE(x) #x"() is not valid on cc70 and above, and should be replaced with "#x"_sync()." #endif __SM_30_INTRINSICS_DECL__ unsigned __fns(unsigned mask, unsigned base, int offset) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ void __barrier_sync(unsigned id) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ void __barrier_sync_count(unsigned id, unsigned cnt) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ void __syncwarp(unsigned mask=0xFFFFFFFF) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __all_sync(unsigned mask, int pred) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __any_sync(unsigned mask, int pred) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __uni_sync(unsigned mask, int pred) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned __ballot_sync(unsigned mask, int pred) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned __activemask() __DEF_IF_HOST #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 700 __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) int __shfl(int var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) unsigned int __shfl(unsigned int var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) int __shfl_up(int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) unsigned int __shfl_up(unsigned int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) int __shfl_down(int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) unsigned int __shfl_down(unsigned int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) int __shfl_xor(int var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) unsigned int __shfl_xor(unsigned int var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) float __shfl(float var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) float __shfl_up(float var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) float __shfl_down(float var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) float __shfl_xor(float var, int laneMask, int width=warpSize) __DEF_IF_HOST #endif __SM_30_INTRINSICS_DECL__ int __shfl_sync(unsigned mask, int var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned int __shfl_sync(unsigned mask, unsigned int var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __shfl_up_sync(unsigned mask, int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned int __shfl_up_sync(unsigned mask, unsigned int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __shfl_down_sync(unsigned mask, int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned int __shfl_down_sync(unsigned mask, unsigned int var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ int __shfl_xor_sync(unsigned mask, int var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned int __shfl_xor_sync(unsigned mask, unsigned int var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ float __shfl_sync(unsigned mask, float var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ float __shfl_up_sync(unsigned mask, float var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ float __shfl_down_sync(unsigned mask, float var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ float __shfl_xor_sync(unsigned mask, float var, int laneMask, int width=warpSize) __DEF_IF_HOST #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 700 __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) unsigned long long __shfl(unsigned long long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) long long __shfl(long long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) long long __shfl_up(long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) unsigned long long __shfl_up(unsigned long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) long long __shfl_down(long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) unsigned long long __shfl_down(unsigned long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) long long __shfl_xor(long long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) unsigned long long __shfl_xor(unsigned long long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) double __shfl(double var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) double __shfl_up(double var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) double __shfl_down(double var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) double __shfl_xor(double var, int laneMask, int width=warpSize) __DEF_IF_HOST #endif __SM_30_INTRINSICS_DECL__ long long __shfl_sync(unsigned mask, long long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long long __shfl_sync(unsigned mask, unsigned long long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long long __shfl_up_sync(unsigned mask, long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long long __shfl_up_sync(unsigned mask, unsigned long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long long __shfl_down_sync(unsigned mask, long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long long __shfl_down_sync(unsigned mask, unsigned long long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long long __shfl_xor_sync(unsigned mask, long long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long long __shfl_xor_sync(unsigned mask, unsigned long long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ double __shfl_sync(unsigned mask, double var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ double __shfl_up_sync(unsigned mask, double var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ double __shfl_down_sync(unsigned mask, double var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ double __shfl_xor_sync(unsigned mask, double var, int laneMask, int width=warpSize) __DEF_IF_HOST #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 700 __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) long __shfl(long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl)) unsigned long __shfl(unsigned long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) long __shfl_up(long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_up)) unsigned long __shfl_up(unsigned long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) long __shfl_down(long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_down)) unsigned long __shfl_down(unsigned long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) long __shfl_xor(long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ __DEPRECATED__(__WSB_DEPRECATION_MESSAGE(__shfl_xor)) unsigned long __shfl_xor(unsigned long var, int laneMask, int width=warpSize) __DEF_IF_HOST #endif __SM_30_INTRINSICS_DECL__ long __shfl_sync(unsigned mask, long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long __shfl_sync(unsigned mask, unsigned long var, int srcLane, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long __shfl_up_sync(unsigned mask, long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long __shfl_up_sync(unsigned mask, unsigned long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long __shfl_down_sync(unsigned mask, long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long __shfl_down_sync(unsigned mask, unsigned long var, unsigned int delta, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ long __shfl_xor_sync(unsigned mask, long var, int laneMask, int width=warpSize) __DEF_IF_HOST __SM_30_INTRINSICS_DECL__ unsigned long __shfl_xor_sync(unsigned mask, unsigned long var, int laneMask, int width=warpSize) __DEF_IF_HOST #undef __DEPRECATED__ #undef __WSB_DEPRECATION_MESSAGE #if defined(__local_warpSize) #undef warpSize #undef __local_warpSize #endif #endif #endif #undef __DEF_IF_HOST #undef __SM_30_INTRINSICS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_30_intrinsics.hpp" #endif #endif #if !defined(__SM_32_INTRINSICS_H__) #define __SM_32_INTRINSICS_H__ #if defined(__CUDACC_RTC__) #define __SM_32_INTRINSICS_DECL__ __device__ #else #define __SM_32_INTRINSICS_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA) #define __DEF_IF_HOST ; #else #define __DEF_IF_HOST { } #endif __SM_32_INTRINSICS_DECL__ long __ldg(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldg(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldg(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldg(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldg(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldg(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldg(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldg(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldg(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldg(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldg(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldg(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldg(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldg(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldg(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldg(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldg(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldg(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldg(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldg(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldg(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldg(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldg(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldg(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldg(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldg(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldg(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldg(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldg(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldg(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcg(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcg(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcg(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcg(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcg(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcg(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcg(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcg(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcg(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcg(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcg(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcg(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcg(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcg(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcg(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcg(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcg(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcg(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcg(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcg(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcg(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcg(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcg(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcg(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcg(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcg(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcg(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcg(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcg(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcg(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldca(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldca(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldca(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldca(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldca(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldca(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldca(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldca(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldca(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldca(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldca(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldca(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldca(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldca(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldca(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldca(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldca(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldca(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldca(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldca(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldca(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldca(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldca(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldca(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldca(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldca(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldca(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldca(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldca(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldca(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcs(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcs(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcs(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcs(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcs(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcs(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcs(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcs(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcs(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcs(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcs(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcs(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcs(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcs(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcs(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcs(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcs(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcs(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcs(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcs(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcs(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcs(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcs(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcs(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcs(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcs(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcs(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcs(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcs(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcs(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldlu(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldlu(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldlu(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldlu(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldlu(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldlu(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldlu(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldlu(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldlu(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldlu(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldlu(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldlu(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldlu(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldlu(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldlu(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldlu(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldlu(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldlu(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldlu(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldlu(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldlu(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldlu(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldlu(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldlu(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldlu(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldlu(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldlu(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldlu(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldlu(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldlu(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcv(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcv(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcv(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcv(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcv(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcv(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcv(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcv(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcv(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcv(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcv(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcv(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcv(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcv(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcv(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcv(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcv(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcv(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcv(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcv(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcv(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcv(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcv(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcv(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcv(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcv(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcv(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcv(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcv(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcv(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_l(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_lc(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_r(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_rc(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST #endif #endif #undef __SM_32_INTRINSICS_DECL__ #if !defined(__CUDACC_RTC__) && (defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA)) #include "sm_32_intrinsics.hpp" #endif #endif #if !defined(__SM_35_INTRINSICS_H__) #define __SM_35_INTRINSICS_H__ #if !defined(__SM_32_INTRINSICS_H__) #define __SM_32_INTRINSICS_H__ #if defined(__CUDACC_RTC__) #define __SM_32_INTRINSICS_DECL__ __device__ #else #define __SM_32_INTRINSICS_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA) #define __DEF_IF_HOST ; #else #define __DEF_IF_HOST { } #endif __SM_32_INTRINSICS_DECL__ long __ldg(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldg(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldg(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldg(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldg(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldg(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldg(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldg(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldg(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldg(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldg(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldg(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldg(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldg(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldg(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldg(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldg(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldg(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldg(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldg(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldg(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldg(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldg(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldg(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldg(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldg(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldg(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldg(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldg(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldg(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcg(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcg(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcg(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcg(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcg(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcg(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcg(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcg(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcg(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcg(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcg(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcg(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcg(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcg(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcg(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcg(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcg(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcg(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcg(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcg(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcg(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcg(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcg(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcg(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcg(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcg(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcg(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcg(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcg(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcg(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldca(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldca(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldca(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldca(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldca(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldca(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldca(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldca(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldca(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldca(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldca(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldca(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldca(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldca(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldca(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldca(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldca(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldca(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldca(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldca(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldca(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldca(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldca(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldca(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldca(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldca(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldca(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldca(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldca(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldca(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcs(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcs(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcs(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcs(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcs(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcs(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcs(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcs(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcs(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcs(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcs(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcs(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcs(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcs(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcs(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcs(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcs(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcs(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcs(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcs(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcs(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcs(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcs(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcs(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcs(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcs(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcs(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcs(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcs(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcs(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldlu(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldlu(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldlu(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldlu(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldlu(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldlu(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldlu(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldlu(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldlu(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldlu(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldlu(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldlu(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldlu(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldlu(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldlu(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldlu(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldlu(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldlu(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldlu(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldlu(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldlu(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldlu(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldlu(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldlu(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldlu(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldlu(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldlu(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldlu(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldlu(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldlu(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long __ldcv(const long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long __ldcv(const unsigned long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char __ldcv(const char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ signed char __ldcv(const signed char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short __ldcv(const short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int __ldcv(const int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ long long __ldcv(const long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char2 __ldcv(const char2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ char4 __ldcv(const char4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short2 __ldcv(const short2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ short4 __ldcv(const short4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int2 __ldcv(const int2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ int4 __ldcv(const int4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ longlong2 __ldcv(const longlong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned char __ldcv(const unsigned char *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned short __ldcv(const unsigned short *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __ldcv(const unsigned int *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned long long __ldcv(const unsigned long long *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar2 __ldcv(const uchar2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uchar4 __ldcv(const uchar4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort2 __ldcv(const ushort2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ushort4 __ldcv(const ushort4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint2 __ldcv(const uint2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ uint4 __ldcv(const uint4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ ulonglong2 __ldcv(const ulonglong2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float __ldcv(const float *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double __ldcv(const double *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float2 __ldcv(const float2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ float4 __ldcv(const float4 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ double2 __ldcv(const double2 *ptr) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwb(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcg(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stcs(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(long *ptr, long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned long *ptr, unsigned long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char *ptr, char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(signed char *ptr, signed char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short *ptr, short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int *ptr, int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(long long *ptr, long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char2 *ptr, char2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(char4 *ptr, char4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short2 *ptr, short2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(short4 *ptr, short4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int2 *ptr, int2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(int4 *ptr, int4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(longlong2 *ptr, longlong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned char *ptr, unsigned char value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned short *ptr, unsigned short value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned int *ptr, unsigned int value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(unsigned long long *ptr, unsigned long long value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uchar2 *ptr, uchar2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uchar4 *ptr, uchar4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ushort2 *ptr, ushort2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ushort4 *ptr, ushort4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uint2 *ptr, uint2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(uint4 *ptr, uint4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(ulonglong2 *ptr, ulonglong2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float *ptr, float value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(double *ptr, double value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float2 *ptr, float2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(float4 *ptr, float4 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ void __stwt(double2 *ptr, double2 value) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_l(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_lc(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_r(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST __SM_32_INTRINSICS_DECL__ unsigned int __funnelshift_rc(unsigned int lo, unsigned int hi, unsigned int shift) __DEF_IF_HOST #endif #endif #undef __SM_32_INTRINSICS_DECL__ #if !defined(__CUDACC_RTC__) && (defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA)) #include "sm_32_intrinsics.hpp" #endif #endif #endif #if !defined(__SM_61_INTRINSICS_H__) #define __SM_61_INTRINSICS_H__ #if defined(__CUDACC_RTC__) #define __SM_61_INTRINSICS_DECL__ __device__ #else #define __SM_61_INTRINSICS_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 610 #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif __SM_61_INTRINSICS_DECL__ int __dp2a_lo(int srcA, int srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp2a_lo(unsigned int srcA, unsigned int srcB, unsigned int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ int __dp2a_lo(short2 srcA, char4 srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp2a_lo(ushort2 srcA, uchar4 srcB, unsigned int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ int __dp2a_hi(int srcA, int srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp2a_hi(unsigned int srcA, unsigned int srcB, unsigned int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ int __dp2a_hi(short2 srcA, char4 srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp2a_hi(ushort2 srcA, uchar4 srcB, unsigned int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ int __dp4a(int srcA, int srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp4a(unsigned int srcA, unsigned int srcB, unsigned int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ int __dp4a(char4 srcA, char4 srcB, int c) __DEF_IF_HOST __SM_61_INTRINSICS_DECL__ unsigned int __dp4a(uchar4 srcA, uchar4 srcB, unsigned int c) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_61_INTRINSICS_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_61_intrinsics.hpp" #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/sm_70_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/sm_70_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_70_RT_H__ #endif #if !defined(__SM_70_RT_H__) #define __SM_70_RT_H__ #if defined(__CUDACC_RTC__) #define __SM_70_RT_DECL__ __host__ __device__ #else #define __SM_70_RT_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 700 #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, int value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, unsigned long value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, long value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, unsigned long long value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, long long value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, float value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_any_sync(unsigned mask, double value) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, unsigned value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, int value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, unsigned long value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, long value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, unsigned long long value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, long long value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, float value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned int __match_all_sync(unsigned mask, double value, int *pred) __DEF_IF_HOST __SM_70_RT_DECL__ void __nanosleep(unsigned int ns) __DEF_IF_HOST __SM_70_RT_DECL__ unsigned short int atomicCAS(unsigned short int *address, unsigned short int compare, unsigned short int val) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_70_RT_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_70_rt.hpp" #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_70_RT_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_70_RT_H__ #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/sm_80_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/sm_80_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_80_RT_H__ #endif #if !defined(__SM_80_RT_H__) #define __SM_80_RT_H__ #if defined(__CUDACC_RTC__) #define __SM_80_RT_DECL__ __host__ __device__ #else #define __SM_80_RT_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 800 #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif __SM_80_RT_DECL__ unsigned __reduce_add_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_80_RT_DECL__ unsigned __reduce_min_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_80_RT_DECL__ unsigned __reduce_max_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_80_RT_DECL__ int __reduce_add_sync(unsigned mask, int value) __DEF_IF_HOST __SM_80_RT_DECL__ int __reduce_min_sync(unsigned mask, int value) __DEF_IF_HOST __SM_80_RT_DECL__ int __reduce_max_sync(unsigned mask, int value) __DEF_IF_HOST __SM_80_RT_DECL__ unsigned __reduce_and_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_80_RT_DECL__ unsigned __reduce_or_sync(unsigned mask, unsigned value) __DEF_IF_HOST __SM_80_RT_DECL__ unsigned __reduce_xor_sync(unsigned mask, unsigned value) __DEF_IF_HOST extern "C" { inline __device__ void *__nv_associate_access_property(const void *ptr, unsigned long long property) { extern __device__ void *__nv_associate_access_property_impl(const void *, unsigned long long); return __nv_associate_access_property_impl(ptr, property); } inline __device__ void __nv_memcpy_async_shared_global_4(void *dst, const void *src, unsigned src_size) { extern __device__ void __nv_memcpy_async_shared_global_4_impl(void *, const void *, unsigned); __nv_memcpy_async_shared_global_4_impl(dst, src, src_size); } inline __device__ void __nv_memcpy_async_shared_global_8(void *dst, const void *src, unsigned src_size) { extern __device__ void __nv_memcpy_async_shared_global_8_impl(void *, const void *, unsigned); __nv_memcpy_async_shared_global_8_impl(dst, src, src_size); } inline __device__ void __nv_memcpy_async_shared_global_16(void *dst, const void *src, unsigned src_size) { extern __device__ void __nv_memcpy_async_shared_global_16_impl(void *, const void *, unsigned); __nv_memcpy_async_shared_global_16_impl(dst, src, src_size); } } #endif #endif #undef __DEF_IF_HOST #undef __SM_80_RT_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_80_rt.hpp" #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_80_RT_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_80_RT_H__ #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/sm_90_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/sm_90_rt.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_90_RT_H__ #endif #if !defined(__SM_90_RT_H__) #define __SM_90_RT_H__ #if defined(__CUDACC_RTC__) #define __SM_90_RT_DECL__ __host__ __device__ #else #define __SM_90_RT_DECL__ static __device__ __inline__ #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 900 #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else #define __DEF_IF_HOST ; #endif __SM_90_RT_DECL__ unsigned __isCtaShared(const void *ptr) __DEF_IF_HOST __SM_90_RT_DECL__ unsigned __isClusterShared(const void *ptr) __DEF_IF_HOST __SM_90_RT_DECL__ void *__cluster_map_shared_rank(const void *ptr, unsigned target_block_rank) __DEF_IF_HOST __SM_90_RT_DECL__ unsigned __cluster_query_shared_rank(const void *ptr) __DEF_IF_HOST __SM_90_RT_DECL__ uint2 __cluster_map_shared_multicast(const void *ptr, unsigned cluster_cta_mask) __DEF_IF_HOST __SM_90_RT_DECL__ unsigned __clusterDimIsSpecified() __DEF_IF_HOST __SM_90_RT_DECL__ dim3 __clusterDim() __DEF_IF_HOST __SM_90_RT_DECL__ dim3 __clusterRelativeBlockIdx() __DEF_IF_HOST __SM_90_RT_DECL__ dim3 __clusterGridDimInClusters() __DEF_IF_HOST __SM_90_RT_DECL__ dim3 __clusterIdx() __DEF_IF_HOST __SM_90_RT_DECL__ unsigned __clusterRelativeBlockRank() __DEF_IF_HOST __SM_90_RT_DECL__ unsigned __clusterSizeInBlocks() __DEF_IF_HOST __SM_90_RT_DECL__ void __cluster_barrier_arrive() __DEF_IF_HOST __SM_90_RT_DECL__ void __cluster_barrier_arrive_relaxed() __DEF_IF_HOST __SM_90_RT_DECL__ void __cluster_barrier_wait() __DEF_IF_HOST __SM_90_RT_DECL__ void __threadfence_cluster() __DEF_IF_HOST __SM_90_RT_DECL__ float2 atomicAdd(float2 *address, float2 val) __DEF_IF_HOST __SM_90_RT_DECL__ float2 atomicAdd_block(float2 *address, float2 val) __DEF_IF_HOST __SM_90_RT_DECL__ float2 atomicAdd_system(float2 *address, float2 val) __DEF_IF_HOST __SM_90_RT_DECL__ float4 atomicAdd(float4 *address, float4 val) __DEF_IF_HOST __SM_90_RT_DECL__ float4 atomicAdd_block(float4 *address, float4 val) __DEF_IF_HOST __SM_90_RT_DECL__ float4 atomicAdd_system(float4 *address, float4 val) __DEF_IF_HOST #endif #endif #undef __DEF_IF_HOST #undef __SM_90_RT_DECL__ #if !defined(__CUDACC_RTC__) && defined(__CUDA_ARCH__) #include "sm_90_rt.hpp" #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_90_RT_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_SM_90_RT_H__ #endif #ifndef __TEXTURE_INDIRECT_FUNCTIONS_H__ #define __TEXTURE_INDIRECT_FUNCTIONS_H__ #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif #if defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 600) #define __NV_TEX_SPARSE 1 #endif template <typename T> struct __nv_itex_trait { }; template<> struct __nv_itex_trait<char> { typedef void type; }; template<> struct __nv_itex_trait<signed char> { typedef void type; }; template<> struct __nv_itex_trait<char1> { typedef void type; }; template<> struct __nv_itex_trait<char2> { typedef void type; }; template<> struct __nv_itex_trait<char4> { typedef void type; }; template<> struct __nv_itex_trait<unsigned char> { typedef void type; }; template<> struct __nv_itex_trait<uchar1> { typedef void type; }; template<> struct __nv_itex_trait<uchar2> { typedef void type; }; template<> struct __nv_itex_trait<uchar4> { typedef void type; }; template<> struct __nv_itex_trait<short> { typedef void type; }; template<> struct __nv_itex_trait<short1> { typedef void type; }; template<> struct __nv_itex_trait<short2> { typedef void type; }; template<> struct __nv_itex_trait<short4> { typedef void type; }; template<> struct __nv_itex_trait<unsigned short> { typedef void type; }; template<> struct __nv_itex_trait<ushort1> { typedef void type; }; template<> struct __nv_itex_trait<ushort2> { typedef void type; }; template<> struct __nv_itex_trait<ushort4> { typedef void type; }; template<> struct __nv_itex_trait<int> { typedef void type; }; template<> struct __nv_itex_trait<int1> { typedef void type; }; template<> struct __nv_itex_trait<int2> { typedef void type; }; template<> struct __nv_itex_trait<int4> { typedef void type; }; template<> struct __nv_itex_trait<unsigned int> { typedef void type; }; template<> struct __nv_itex_trait<uint1> { typedef void type; }; template<> struct __nv_itex_trait<uint2> { typedef void type; }; template<> struct __nv_itex_trait<uint4> { typedef void type; }; #if !defined(__LP64__) template<> struct __nv_itex_trait<long> { typedef void type; }; template<> struct __nv_itex_trait<long1> { typedef void type; }; template<> struct __nv_itex_trait<long2> { typedef void type; }; template<> struct __nv_itex_trait<long4> { typedef void type; }; template<> struct __nv_itex_trait<unsigned long> { typedef void type; }; template<> struct __nv_itex_trait<ulong1> { typedef void type; }; template<> struct __nv_itex_trait<ulong2> { typedef void type; }; template<> struct __nv_itex_trait<ulong4> { typedef void type; }; #endif template<> struct __nv_itex_trait<float> { typedef void type; }; template<> struct __nv_itex_trait<float1> { typedef void type; }; template<> struct __nv_itex_trait<float2> { typedef void type; }; template<> struct __nv_itex_trait<float4> { typedef void type; }; template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1Dfetch(T *ptr, cudaTextureObject_t obj, int x) { __nv_tex_surf_handler("__itex1Dfetch", ptr, obj, x); } template <class T> static __device__ T tex1Dfetch(cudaTextureObject_t texObject, int x) { T ret; tex1Dfetch(&ret, texObject, x); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1D(T *ptr, cudaTextureObject_t obj, float x) { __nv_tex_surf_handler("__itex1D", ptr, obj, x); } template <class T> static __device__ T tex1D(cudaTextureObject_t texObject, float x) { T ret; tex1D(&ret, texObject, x); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2D(T *ptr, cudaTextureObject_t obj, float x, float y) { __nv_tex_surf_handler("__itex2D", ptr, obj, x, y); } template <class T> static __device__ T tex2D(cudaTextureObject_t texObject, float x, float y) { T ret; tex2D(&ret, texObject, x, y); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2D(T *ptr, cudaTextureObject_t obj, float x, float y, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2D_sparse", ptr, obj, x, y, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2D(cudaTextureObject_t texObject, float x, float y, bool* isResident) { T ret; tex2D(&ret, texObject, x, y, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3D(T *ptr, cudaTextureObject_t obj, float x, float y, float z) { __nv_tex_surf_handler("__itex3D", ptr, obj, x, y, z); } template <class T> static __device__ T tex3D(cudaTextureObject_t texObject, float x, float y, float z) { T ret; tex3D(&ret, texObject, x, y, z); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3D(T *ptr, cudaTextureObject_t obj, float x, float y, float z, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex3D_sparse", ptr, obj, x, y, z, &res); *isResident = (res != 0); } template <class T> static __device__ T tex3D(cudaTextureObject_t texObject, float x, float y, float z, bool* isResident) { T ret; tex3D(&ret, texObject, x, y, z, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1DLayered(T *ptr, cudaTextureObject_t obj, float x, int layer) { __nv_tex_surf_handler("__itex1DLayered", ptr, obj, x, layer); } template <class T> static __device__ T tex1DLayered(cudaTextureObject_t texObject, float x, int layer) { T ret; tex1DLayered(&ret, texObject, x, layer); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayered(T *ptr, cudaTextureObject_t obj, float x, float y, int layer) { __nv_tex_surf_handler("__itex2DLayered", ptr, obj, x, y, layer); } template <class T> static __device__ T tex2DLayered(cudaTextureObject_t texObject, float x, float y, int layer) { T ret; tex2DLayered(&ret, texObject, x, y, layer); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayered(T *ptr, cudaTextureObject_t obj, float x, float y, int layer, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2DLayered_sparse", ptr, obj, x, y, layer, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2DLayered(cudaTextureObject_t texObject, float x, float y, int layer, bool* isResident) { T ret; tex2DLayered(&ret, texObject, x, y, layer, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemap(T *ptr, cudaTextureObject_t obj, float x, float y, float z) { __nv_tex_surf_handler("__itexCubemap", ptr, obj, x, y, z); } template <class T> static __device__ T texCubemap(cudaTextureObject_t texObject, float x, float y, float z) { T ret; texCubemap(&ret, texObject, x, y, z); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemapLayered(T *ptr, cudaTextureObject_t obj, float x, float y, float z, int layer) { __nv_tex_surf_handler("__itexCubemapLayered", ptr, obj, x, y, z, layer); } template <class T> static __device__ T texCubemapLayered(cudaTextureObject_t texObject, float x, float y, float z, int layer) { T ret; texCubemapLayered(&ret, texObject, x, y, z, layer); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2Dgather(T *ptr, cudaTextureObject_t obj, float x, float y, int comp = 0) { __nv_tex_surf_handler("__itex2Dgather", ptr, obj, x, y, comp); } template <class T> static __device__ T tex2Dgather(cudaTextureObject_t to, float x, float y, int comp = 0) { T ret; tex2Dgather(&ret, to, x, y, comp); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2Dgather(T *ptr, cudaTextureObject_t obj, float x, float y, bool* isResident, int comp = 0) { unsigned char res; __nv_tex_surf_handler("__itex2Dgather_sparse", ptr, obj, x, y, comp, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2Dgather(cudaTextureObject_t to, float x, float y, bool* isResident, int comp = 0) { T ret; tex2Dgather(&ret, to, x, y, isResident, comp); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1DLod(T *ptr, cudaTextureObject_t obj, float x, float level) { __nv_tex_surf_handler("__itex1DLod", ptr, obj, x, level); } template <class T> static __device__ T tex1DLod(cudaTextureObject_t texObject, float x, float level) { T ret; tex1DLod(&ret, texObject, x, level); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLod(T *ptr, cudaTextureObject_t obj, float x, float y, float level) { __nv_tex_surf_handler("__itex2DLod", ptr, obj, x, y, level); } template <class T> static __device__ T tex2DLod(cudaTextureObject_t texObject, float x, float y, float level) { T ret; tex2DLod(&ret, texObject, x, y, level); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLod(T *ptr, cudaTextureObject_t obj, float x, float y, float level, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2DLod_sparse", ptr, obj, x, y, level, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2DLod(cudaTextureObject_t texObject, float x, float y, float level, bool* isResident) { T ret; tex2DLod(&ret, texObject, x, y, level, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3DLod(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float level) { __nv_tex_surf_handler("__itex3DLod", ptr, obj, x, y, z, level); } template <class T> static __device__ T tex3DLod(cudaTextureObject_t texObject, float x, float y, float z, float level) { T ret; tex3DLod(&ret, texObject, x, y, z, level); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3DLod(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float level, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex3DLod_sparse", ptr, obj, x, y, z, level, &res); *isResident = (res != 0); } template <class T> static __device__ T tex3DLod(cudaTextureObject_t texObject, float x, float y, float z, float level, bool* isResident) { T ret; tex3DLod(&ret, texObject, x, y, z, level, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1DLayeredLod(T *ptr, cudaTextureObject_t obj, float x, int layer, float level) { __nv_tex_surf_handler("__itex1DLayeredLod", ptr, obj, x, layer, level); } template <class T> static __device__ T tex1DLayeredLod(cudaTextureObject_t texObject, float x, int layer, float level) { T ret; tex1DLayeredLod(&ret, texObject, x, layer, level); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayeredLod(T *ptr, cudaTextureObject_t obj, float x, float y, int layer, float level) { __nv_tex_surf_handler("__itex2DLayeredLod", ptr, obj, x, y, layer, level); } template <class T> static __device__ T tex2DLayeredLod(cudaTextureObject_t texObject, float x, float y, int layer, float level) { T ret; tex2DLayeredLod(&ret, texObject, x, y, layer, level); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayeredLod(T *ptr, cudaTextureObject_t obj, float x, float y, int layer, float level, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2DLayeredLod_sparse", ptr, obj, x, y, layer, level, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2DLayeredLod(cudaTextureObject_t texObject, float x, float y, int layer, float level, bool* isResident) { T ret; tex2DLayeredLod(&ret, texObject, x, y, layer, level, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemapLod(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float level) { __nv_tex_surf_handler("__itexCubemapLod", ptr, obj, x, y, z, level); } template <class T> static __device__ T texCubemapLod(cudaTextureObject_t texObject, float x, float y, float z, float level) { T ret; texCubemapLod(&ret, texObject, x, y, z, level); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemapGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float4 dPdx, float4 dPdy) { __nv_tex_surf_handler("__itexCubemapGrad_v2", ptr, obj, x, y, z, &dPdx, &dPdy); } template <class T> static __device__ T texCubemapGrad(cudaTextureObject_t texObject, float x, float y, float z, float4 dPdx, float4 dPdy) { T ret; texCubemapGrad(&ret, texObject, x, y, z, dPdx, dPdy); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemapLayeredLod(T *ptr, cudaTextureObject_t obj, float x, float y, float z, int layer, float level) { __nv_tex_surf_handler("__itexCubemapLayeredLod", ptr, obj, x, y, z, layer, level); } template <class T> static __device__ T texCubemapLayeredLod(cudaTextureObject_t texObject, float x, float y, float z, int layer, float level) { T ret; texCubemapLayeredLod(&ret, texObject, x, y, z, layer, level); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1DGrad(T *ptr, cudaTextureObject_t obj, float x, float dPdx, float dPdy) { __nv_tex_surf_handler("__itex1DGrad", ptr, obj, x, dPdx, dPdy); } template <class T> static __device__ T tex1DGrad(cudaTextureObject_t texObject, float x, float dPdx, float dPdy) { T ret; tex1DGrad(&ret, texObject, x, dPdx, dPdy); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float2 dPdx, float2 dPdy) { __nv_tex_surf_handler("__itex2DGrad_v2", ptr, obj, x, y, &dPdx, &dPdy); } template <class T> static __device__ T tex2DGrad(cudaTextureObject_t texObject, float x, float y, float2 dPdx, float2 dPdy) { T ret; tex2DGrad(&ret, texObject, x, y, dPdx, dPdy); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float2 dPdx, float2 dPdy, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2DGrad_sparse", ptr, obj, x, y, &dPdx, &dPdy, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2DGrad(cudaTextureObject_t texObject, float x, float y, float2 dPdx, float2 dPdy, bool* isResident) { T ret; tex2DGrad(&ret, texObject, x, y, dPdx, dPdy, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3DGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float4 dPdx, float4 dPdy) { __nv_tex_surf_handler("__itex3DGrad_v2", ptr, obj, x, y, z, &dPdx, &dPdy); } template <class T> static __device__ T tex3DGrad(cudaTextureObject_t texObject, float x, float y, float z, float4 dPdx, float4 dPdy) { T ret; tex3DGrad(&ret, texObject, x, y, z, dPdx, dPdy); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex3DGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float z, float4 dPdx, float4 dPdy, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex3DGrad_sparse", ptr, obj, x, y, z, &dPdx, &dPdy, &res); *isResident = (res != 0); } template <class T> static __device__ T tex3DGrad(cudaTextureObject_t texObject, float x, float y, float z, float4 dPdx, float4 dPdy, bool* isResident) { T ret; tex3DGrad(&ret, texObject, x, y, z, dPdx, dPdy, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type tex1DLayeredGrad(T *ptr, cudaTextureObject_t obj, float x, int layer, float dPdx, float dPdy) { __nv_tex_surf_handler("__itex1DLayeredGrad", ptr, obj, x, layer, dPdx, dPdy); } template <class T> static __device__ T tex1DLayeredGrad(cudaTextureObject_t texObject, float x, int layer, float dPdx, float dPdy) { T ret; tex1DLayeredGrad(&ret, texObject, x, layer, dPdx, dPdy); return ret; } template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayeredGrad(T * ptr, cudaTextureObject_t obj, float x, float y, int layer, float2 dPdx, float2 dPdy) { __nv_tex_surf_handler("__itex2DLayeredGrad_v2", ptr, obj, x, y, layer, &dPdx, &dPdy); } template <class T> static __device__ T tex2DLayeredGrad(cudaTextureObject_t texObject, float x, float y, int layer, float2 dPdx, float2 dPdy) { T ret; tex2DLayeredGrad(&ret, texObject, x, y, layer, dPdx, dPdy); return ret; } #if __NV_TEX_SPARSE template <typename T> static __device__ typename __nv_itex_trait<T>::type tex2DLayeredGrad(T * ptr, cudaTextureObject_t obj, float x, float y, int layer, float2 dPdx, float2 dPdy, bool* isResident) { unsigned char res; __nv_tex_surf_handler("__itex2DLayeredGrad_sparse", ptr, obj, x, y, layer, &dPdx, &dPdy, &res); *isResident = (res != 0); } template <class T> static __device__ T tex2DLayeredGrad(cudaTextureObject_t texObject, float x, float y, int layer, float2 dPdx, float2 dPdy, bool* isResident) { T ret; tex2DLayeredGrad(&ret, texObject, x, y, layer, dPdx, dPdy, isResident); return ret; } #endif template <typename T> static __device__ typename __nv_itex_trait<T>::type texCubemapLayeredGrad(T *ptr, cudaTextureObject_t obj, float x, float y, float z, int layer, float4 dPdx, float4 dPdy) { __nv_tex_surf_handler("__itexCubemapLayeredGrad_v2", ptr, obj, x, y, z, layer, &dPdx, &dPdy); } template <class T> static __device__ T texCubemapLayeredGrad(cudaTextureObject_t texObject, float x, float y, float z, int layer, float4 dPdx, float4 dPdy) { T ret; texCubemapLayeredGrad(&ret, texObject, x, y, z, layer, dPdx, dPdy); return ret; } #undef __NV_TEX_SPARSE #endif #endif #ifndef __SURFACE_INDIRECT_FUNCTIONS_H__ #define __SURFACE_INDIRECT_FUNCTIONS_H__ #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_RUNTIME_API_H__) #define __CUDA_RUNTIME_API_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #define CUDART_VERSION 12010 #if defined(__CUDA_API_VER_MAJOR__) && defined(__CUDA_API_VER_MINOR__) # define __CUDART_API_VERSION ((__CUDA_API_VER_MAJOR__ * 1000) + (__CUDA_API_VER_MINOR__ * 10)) #else # define __CUDART_API_VERSION CUDART_VERSION #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) || defined(__CUDA_API_VERSION_INTERNAL) #define __CUDART_API_PER_THREAD_DEFAULT_STREAM #define __CUDART_API_PTDS(api) api ## _ptds #define __CUDART_API_PTSZ(api) api ## _ptsz #else #define __CUDART_API_PTDS(api) api #define __CUDART_API_PTSZ(api) api #endif #define cudaSignalExternalSemaphoresAsync __CUDART_API_PTSZ(cudaSignalExternalSemaphoresAsync_v2) #define cudaWaitExternalSemaphoresAsync __CUDART_API_PTSZ(cudaWaitExternalSemaphoresAsync_v2) #define cudaStreamGetCaptureInfo __CUDART_API_PTSZ(cudaStreamGetCaptureInfo_v2) #define cudaGetDeviceProperties cudaGetDeviceProperties_v2 #if defined(__CUDART_API_PER_THREAD_DEFAULT_STREAM) #define cudaMemcpy __CUDART_API_PTDS(cudaMemcpy) #define cudaMemcpyToSymbol __CUDART_API_PTDS(cudaMemcpyToSymbol) #define cudaMemcpyFromSymbol __CUDART_API_PTDS(cudaMemcpyFromSymbol) #define cudaMemcpy2D __CUDART_API_PTDS(cudaMemcpy2D) #define cudaMemcpyToArray __CUDART_API_PTDS(cudaMemcpyToArray) #define cudaMemcpy2DToArray __CUDART_API_PTDS(cudaMemcpy2DToArray) #define cudaMemcpyFromArray __CUDART_API_PTDS(cudaMemcpyFromArray) #define cudaMemcpy2DFromArray __CUDART_API_PTDS(cudaMemcpy2DFromArray) #define cudaMemcpyArrayToArray __CUDART_API_PTDS(cudaMemcpyArrayToArray) #define cudaMemcpy2DArrayToArray __CUDART_API_PTDS(cudaMemcpy2DArrayToArray) #define cudaMemcpy3D __CUDART_API_PTDS(cudaMemcpy3D) #define cudaMemcpy3DPeer __CUDART_API_PTDS(cudaMemcpy3DPeer) #define cudaMemset __CUDART_API_PTDS(cudaMemset) #define cudaMemset2D __CUDART_API_PTDS(cudaMemset2D) #define cudaMemset3D __CUDART_API_PTDS(cudaMemset3D) #define cudaGraphInstantiateWithParams __CUDART_API_PTSZ(cudaGraphInstantiateWithParams) #define cudaGraphUpload __CUDART_API_PTSZ(cudaGraphUpload) #define cudaGraphLaunch __CUDART_API_PTSZ(cudaGraphLaunch) #define cudaStreamBeginCapture __CUDART_API_PTSZ(cudaStreamBeginCapture) #define cudaStreamEndCapture __CUDART_API_PTSZ(cudaStreamEndCapture) #define cudaStreamIsCapturing __CUDART_API_PTSZ(cudaStreamIsCapturing) #define cudaMemcpyAsync __CUDART_API_PTSZ(cudaMemcpyAsync) #define cudaMemcpyToSymbolAsync __CUDART_API_PTSZ(cudaMemcpyToSymbolAsync) #define cudaMemcpyFromSymbolAsync __CUDART_API_PTSZ(cudaMemcpyFromSymbolAsync) #define cudaMemcpy2DAsync __CUDART_API_PTSZ(cudaMemcpy2DAsync) #define cudaMemcpyToArrayAsync __CUDART_API_PTSZ(cudaMemcpyToArrayAsync) #define cudaMemcpy2DToArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DToArrayAsync) #define cudaMemcpyFromArrayAsync __CUDART_API_PTSZ(cudaMemcpyFromArrayAsync) #define cudaMemcpy2DFromArrayAsync __CUDART_API_PTSZ(cudaMemcpy2DFromArrayAsync) #define cudaMemcpy3DAsync __CUDART_API_PTSZ(cudaMemcpy3DAsync) #define cudaMemcpy3DPeerAsync __CUDART_API_PTSZ(cudaMemcpy3DPeerAsync) #define cudaMemsetAsync __CUDART_API_PTSZ(cudaMemsetAsync) #define cudaMemset2DAsync __CUDART_API_PTSZ(cudaMemset2DAsync) #define cudaMemset3DAsync __CUDART_API_PTSZ(cudaMemset3DAsync) #define cudaStreamQuery __CUDART_API_PTSZ(cudaStreamQuery) #define cudaStreamGetFlags __CUDART_API_PTSZ(cudaStreamGetFlags) #define cudaStreamGetId __CUDART_API_PTSZ(cudaStreamGetId) #define cudaStreamGetPriority __CUDART_API_PTSZ(cudaStreamGetPriority) #define cudaEventRecord __CUDART_API_PTSZ(cudaEventRecord) #define cudaEventRecordWithFlags __CUDART_API_PTSZ(cudaEventRecordWithFlags) #define cudaStreamWaitEvent __CUDART_API_PTSZ(cudaStreamWaitEvent) #define cudaStreamAddCallback __CUDART_API_PTSZ(cudaStreamAddCallback) #define cudaStreamAttachMemAsync __CUDART_API_PTSZ(cudaStreamAttachMemAsync) #define cudaStreamSynchronize __CUDART_API_PTSZ(cudaStreamSynchronize) #define cudaLaunchKernel __CUDART_API_PTSZ(cudaLaunchKernel) #define cudaLaunchKernelExC __CUDART_API_PTSZ(cudaLaunchKernelExC) #define cudaLaunchHostFunc __CUDART_API_PTSZ(cudaLaunchHostFunc) #define cudaMemPrefetchAsync __CUDART_API_PTSZ(cudaMemPrefetchAsync) #define cudaLaunchCooperativeKernel __CUDART_API_PTSZ(cudaLaunchCooperativeKernel) #define cudaStreamCopyAttributes __CUDART_API_PTSZ(cudaStreamCopyAttributes) #define cudaStreamGetAttribute __CUDART_API_PTSZ(cudaStreamGetAttribute) #define cudaStreamSetAttribute __CUDART_API_PTSZ(cudaStreamSetAttribute) #define cudaMallocAsync __CUDART_API_PTSZ(cudaMallocAsync) #define cudaFreeAsync __CUDART_API_PTSZ(cudaFreeAsync) #define cudaMallocFromPoolAsync __CUDART_API_PTSZ(cudaMallocFromPoolAsync) #define cudaGetDriverEntryPoint __CUDART_API_PTSZ(cudaGetDriverEntryPoint) #endif #if !defined(__dv) #if defined(__cplusplus) #define __dv(v) \ = v #else #define __dv(v) #endif #endif #if (defined(_NVHPC_CUDA) || !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350)) #define CUDART_DEVICE __device__ #else #define CUDART_DEVICE #endif #if !defined(__CUDACC_RTC__) #endif #undef __dv #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_API_H__ #endif #endif template<typename T> struct __nv_isurf_trait { }; template<> struct __nv_isurf_trait<char> { typedef void type; }; template<> struct __nv_isurf_trait<signed char> { typedef void type; }; template<> struct __nv_isurf_trait<char1> { typedef void type; }; template<> struct __nv_isurf_trait<unsigned char> { typedef void type; }; template<> struct __nv_isurf_trait<uchar1> { typedef void type; }; template<> struct __nv_isurf_trait<short> { typedef void type; }; template<> struct __nv_isurf_trait<short1> { typedef void type; }; template<> struct __nv_isurf_trait<unsigned short> { typedef void type; }; template<> struct __nv_isurf_trait<ushort1> { typedef void type; }; template<> struct __nv_isurf_trait<int> { typedef void type; }; template<> struct __nv_isurf_trait<int1> { typedef void type; }; template<> struct __nv_isurf_trait<unsigned int> { typedef void type; }; template<> struct __nv_isurf_trait<uint1> { typedef void type; }; template<> struct __nv_isurf_trait<long long> { typedef void type; }; template<> struct __nv_isurf_trait<longlong1> { typedef void type; }; template<> struct __nv_isurf_trait<unsigned long long> { typedef void type; }; template<> struct __nv_isurf_trait<ulonglong1> { typedef void type; }; template<> struct __nv_isurf_trait<float> { typedef void type; }; template<> struct __nv_isurf_trait<float1> { typedef void type; }; template<> struct __nv_isurf_trait<char2> { typedef void type; }; template<> struct __nv_isurf_trait<uchar2> { typedef void type; }; template<> struct __nv_isurf_trait<short2> { typedef void type; }; template<> struct __nv_isurf_trait<ushort2> { typedef void type; }; template<> struct __nv_isurf_trait<int2> { typedef void type; }; template<> struct __nv_isurf_trait<uint2> { typedef void type; }; template<> struct __nv_isurf_trait<longlong2> { typedef void type; }; template<> struct __nv_isurf_trait<ulonglong2> { typedef void type; }; template<> struct __nv_isurf_trait<float2> { typedef void type; }; template<> struct __nv_isurf_trait<char4> { typedef void type; }; template<> struct __nv_isurf_trait<uchar4> { typedef void type; }; template<> struct __nv_isurf_trait<short4> { typedef void type; }; template<> struct __nv_isurf_trait<ushort4> { typedef void type; }; template<> struct __nv_isurf_trait<int4> { typedef void type; }; template<> struct __nv_isurf_trait<uint4> { typedef void type; }; template<> struct __nv_isurf_trait<float4> { typedef void type; }; template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf1Dread(T *ptr, cudaSurfaceObject_t obj, int x, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf1Dread", ptr, obj, x, mode); } template <class T> static __device__ T surf1Dread(cudaSurfaceObject_t surfObject, int x, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surf1Dread(&ret, surfObject, x, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf2Dread(T *ptr, cudaSurfaceObject_t obj, int x, int y, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf2Dread", ptr, obj, x, y, mode); } template <class T> static __device__ T surf2Dread(cudaSurfaceObject_t surfObject, int x, int y, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surf2Dread(&ret, surfObject, x, y, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf3Dread(T *ptr, cudaSurfaceObject_t obj, int x, int y, int z, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf3Dread", ptr, obj, x, y, z, mode); } template <class T> static __device__ T surf3Dread(cudaSurfaceObject_t surfObject, int x, int y, int z, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surf3Dread(&ret, surfObject, x, y, z, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf1DLayeredread(T *ptr, cudaSurfaceObject_t obj, int x, int layer, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf1DLayeredread", ptr, obj, x, layer, mode); } template <class T> static __device__ T surf1DLayeredread(cudaSurfaceObject_t surfObject, int x, int layer, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surf1DLayeredread(&ret, surfObject, x, layer, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf2DLayeredread(T *ptr, cudaSurfaceObject_t obj, int x, int y, int layer, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf2DLayeredread", ptr, obj, x, y, layer, mode); } template <class T> static __device__ T surf2DLayeredread(cudaSurfaceObject_t surfObject, int x, int y, int layer, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surf2DLayeredread(&ret, surfObject, x, y, layer, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surfCubemapread(T *ptr, cudaSurfaceObject_t obj, int x, int y, int face, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurfCubemapread", ptr, obj, x, y, face, mode); } template <class T> static __device__ T surfCubemapread(cudaSurfaceObject_t surfObject, int x, int y, int face, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surfCubemapread(&ret, surfObject, x, y, face, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surfCubemapLayeredread(T *ptr, cudaSurfaceObject_t obj, int x, int y, int layerface, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurfCubemapLayeredread", ptr, obj, x, y, layerface, mode); } template <class T> static __device__ T surfCubemapLayeredread(cudaSurfaceObject_t surfObject, int x, int y, int layerface, cudaSurfaceBoundaryMode boundaryMode = cudaBoundaryModeTrap) { T ret; surfCubemapLayeredread(&ret, surfObject, x, y, layerface, boundaryMode); return ret; } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf1Dwrite(T val, cudaSurfaceObject_t obj, int x, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf1Dwrite_v2", &val, obj, x, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf2Dwrite(T val, cudaSurfaceObject_t obj, int x, int y, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf2Dwrite_v2", &val, obj, x, y, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf3Dwrite(T val, cudaSurfaceObject_t obj, int x, int y, int z, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf3Dwrite_v2", &val, obj, x, y, z, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf1DLayeredwrite(T val, cudaSurfaceObject_t obj, int x, int layer, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf1DLayeredwrite_v2", &val, obj, x, layer, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surf2DLayeredwrite(T val, cudaSurfaceObject_t obj, int x, int y, int layer, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurf2DLayeredwrite_v2", &val, obj, x, y, layer, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surfCubemapwrite(T val, cudaSurfaceObject_t obj, int x, int y, int face, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurfCubemapwrite_v2", &val, obj, x, y, face, mode); } template <typename T> static __device__ typename __nv_isurf_trait<T>::type surfCubemapLayeredwrite(T val, cudaSurfaceObject_t obj, int x, int y, int layerface, cudaSurfaceBoundaryMode mode = cudaBoundaryModeTrap) { __nv_tex_surf_handler("__isurfCubemapLayeredwrite_v2", &val, obj, x, y, layerface, mode); } #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/cudacc_ext.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/cudacc_ext.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDACC_EXT_H__ #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDACC_EXT_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDACC_EXT_H__ #endif #ifdef __CUDACC__ extern "C" __host__ __device__ unsigned CUDARTAPI __cudaPushCallConfiguration(dim3 gridDim, dim3 blockDim, size_t sharedMem = 0, struct CUstream_st *stream = 0); #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_FUNCTIONS_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_FUNCTIONS_H__ #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/common_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/common_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_COMMON_FUNCTIONS_H__ #endif #if !defined(__COMMON_FUNCTIONS_H__) #define __COMMON_FUNCTIONS_H__ #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define __CUDACC_VER__ "__CUDACC_VER__ is no longer supported. Use __CUDACC_VER_MAJOR__, __CUDACC_VER_MINOR__, and __CUDACC_VER_BUILD__ instead." #ifndef __CUDA_API_VER_MAJOR__ #define __CUDA_API_VER_MAJOR__ __CUDACC_VER_MAJOR__ #endif #ifndef __CUDA_API_VER_MINOR__ #define __CUDA_API_VER_MINOR__ __CUDACC_VER_MINOR__ #endif #if !defined(__CUDACC_RTC__) #include <string.h> #include <time.h> extern "C" { #endif extern _CRTIMP __host__ __device__ __device_builtin__ __cudart_builtin__ clock_t __cdecl clock(void) #if defined(__QNX__) asm("clock32") #endif __THROW; extern __host__ __device__ __device_builtin__ __cudart_builtin__ void* __cdecl memset(void*, int, size_t) __THROW; extern __host__ __device__ __device_builtin__ __cudart_builtin__ void* __cdecl memcpy(void*, const void*, size_t) __THROW; #if !defined(__CUDACC_RTC__) } #endif #if defined(__CUDA_ARCH__) #if defined(__CUDACC_RTC__) inline __host__ __device__ void* operator new(size_t, void *p) { return p; } inline __host__ __device__ void* operator new[](size_t, void *p) { return p; } inline __host__ __device__ void operator delete(void*, void*) { } inline __host__ __device__ void operator delete[](void*, void*) { } #else #ifndef __CUDA_INTERNAL_SKIP_CPP_HEADERS__ #include <new> #endif #if defined (__GNUC__) #define STD \ std:: #else #define STD #endif extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new(STD size_t, void*) throw(); extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new[](STD size_t, void*) throw(); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*, void*) throw(); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*, void*) throw(); # if __cplusplus >= 201402L || (defined(_MSC_VER) && _MSC_VER >= 1900) || defined(__CUDA_XLC_CPP14__) || defined(__CUDA_ICC_CPP14__) extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*, STD size_t) throw(); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*, STD size_t) throw(); #endif #endif #if !defined(__CUDACC_RTC__) #include <stdio.h> #include <stdlib.h> #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern "C" { extern #if !defined(_MSC_VER) || _MSC_VER < 1900 _CRTIMP #endif #if defined(__GLIBC__) && defined(__GLIBC_MINOR__) && ( (__GLIBC__ < 2) || ( (__GLIBC__ == 2) && (__GLIBC_MINOR__ < 3) ) ) __host__ __device__ __device_builtin__ __cudart_builtin__ int __cdecl printf(const char*, ...) __THROW; #else __host__ __device__ __device_builtin__ __cudart_builtin__ int __cdecl printf(const char*, ...); #endif extern _CRTIMP __host__ __device__ __cudart_builtin__ void* __cdecl malloc(size_t) __THROW; extern _CRTIMP __host__ __device__ __cudart_builtin__ void __cdecl free(void*) __THROW; #if defined(_MSC_VER) extern __host__ __device__ __cudart_builtin__ void* __cdecl _alloca(size_t); #endif #if defined(__QNX__) #undef alloca #define alloca(__S) __builtin_alloca(__S) #endif } #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif #if !defined(__CUDACC_RTC__) #include <assert.h> #endif extern "C" { #if defined(__CUDACC_RTC__) extern __host__ __device__ void __assertfail(const char * __assertion, const char *__file, unsigned int __line, const char *__function, size_t charsize); #elif defined(__APPLE__) #define __builtin_expect(exp,c) (exp) extern __host__ __device__ __cudart_builtin__ void __assert_rtn( const char *, const char *, int, const char *); #elif defined(__ANDROID__) extern __host__ __device__ __cudart_builtin__ void __assert2( const char *, int, const char *, const char *); #elif defined(__QNX__) #if !defined(_LIBCPP_VERSION) namespace std { #endif extern __host__ __device__ __cudart_builtin__ void __assert( const char *, const char *, unsigned int, const char *); #if !defined(_LIBCPP_VERSION) } #endif #elif defined(__HORIZON__) extern __host__ __device__ __cudart_builtin__ void __assert_fail( const char *, const char *, int, const char *); #elif defined(__GNUC__) extern __host__ __device__ __cudart_builtin__ void __assert_fail( const char *, const char *, unsigned int, const char *) __THROW; #elif defined(_WIN32) extern __host__ __device__ __cudart_builtin__ _CRTIMP void __cdecl _wassert( const wchar_t *, const wchar_t *, unsigned); #endif } #if defined(__CUDACC_RTC__) #ifdef NDEBUG #define assert(e) (static_cast<void>(0)) #else #define __ASSERT_STR_HELPER(x) #x #define assert(e) ((e) ? static_cast<void>(0)\ : __assertfail(__ASSERT_STR_HELPER(e), __FILE__,\ __LINE__, __PRETTY_FUNCTION__,\ sizeof(char))) #endif __host__ __device__ void* operator new(size_t); __host__ __device__ void* operator new[](size_t); __host__ __device__ void operator delete(void*); __host__ __device__ void operator delete[](void*); # if __cplusplus >= 201402L __host__ __device__ void operator delete(void*, size_t); __host__ __device__ void operator delete[](void*, size_t); #endif #if __cplusplus >= 201703L namespace std { enum class align_val_t : size_t {}; } __host__ __device__ void* __cdecl operator new(size_t sz, std::align_val_t) noexcept; __host__ __device__ void* __cdecl operator new[](size_t sz, std::align_val_t) noexcept; __host__ __device__ void __cdecl operator delete(void* ptr, std::align_val_t) noexcept; __host__ __device__ void __cdecl operator delete[](void* ptr, std::align_val_t) noexcept; __host__ __device__ void __cdecl operator delete(void* ptr, size_t, std::align_val_t) noexcept; __host__ __device__ void __cdecl operator delete[](void* ptr, size_t, std::align_val_t) noexcept; #endif #else #if defined (__GNUC__) #define __NV_GLIBCXX_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) #if (__cplusplus >= 201103L) && ((!(defined(__QNX__) && defined(_LIBCPP_VERSION))) || (defined(__QNX__) && __NV_GLIBCXX_VERSION >= 80300)) #define THROWBADALLOC #else #if defined(__ANDROID__) && !defined(_LIBCPP_VERSION) && (defined(__BIONIC__) || __NV_GLIBCXX_VERSION < 40900) #define THROWBADALLOC #else #define THROWBADALLOC throw(STD bad_alloc) #endif #endif #define __DELETE_THROW throw() #undef __NV_GLIBCXX_VERSION #else #define THROWBADALLOC throw(...) #endif extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new(STD size_t) THROWBADALLOC; extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new[](STD size_t) THROWBADALLOC; extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*) throw(); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*) throw(); # if __cplusplus >= 201402L || (defined(_MSC_VER) && _MSC_VER >= 1900) || defined(__CUDA_XLC_CPP14__) || defined(__CUDA_ICC_CPP14__) extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*, STD size_t) throw(); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*, STD size_t) throw(); #endif #if __cpp_aligned_new extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new(STD size_t, std::align_val_t); extern __host__ __device__ __cudart_builtin__ void* __cdecl operator new[](STD size_t, std::align_val_t); extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*, std::align_val_t) noexcept; extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*, std::align_val_t) noexcept; extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete(void*, STD size_t, std::align_val_t) noexcept; extern __host__ __device__ __cudart_builtin__ void __cdecl operator delete[](void*, STD size_t, std::align_val_t) noexcept; #endif #undef THROWBADALLOC #undef STD #endif #endif #endif #if defined(__CUDACC_RTC__) && (__CUDA_ARCH__ >= 350) #if !defined(__CUDA_DEVICE_RUNTIME_API_H__) #define __CUDA_DEVICE_RUNTIME_API_H__ #if defined(__CUDACC__) && !defined(__CUDACC_RTC__) #include <stdlib.h> #endif #if !defined(CUDA_FORCE_CDP1_IF_SUPPORTED) && !defined(__CUDADEVRT_INTERNAL__) && !defined(_NVHPC_CUDA) && !(defined(_WIN32) && !defined(_WIN64)) #define __CUDA_INTERNAL_USE_CDP2 #endif #if !defined(__CUDACC_RTC__) #if !defined(__CUDACC_INTERNAL_NO_STUBS__) && !defined(__CUDACC_RDC__) && !defined(__CUDACC_EWP__) && defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 350) && !defined(__CUDADEVRT_INTERNAL__) #if defined(__cplusplus) extern "C" { #endif struct cudaFuncAttributes; #ifndef __CUDA_INTERNAL_USE_CDP2 inline __device__ cudaError_t CUDARTAPI cudaMalloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #else inline __device__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **p, size_t s) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *p, const void *c) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return cudaErrorUnknown; } inline __device__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return cudaErrorUnknown; } #endif #if defined(__cplusplus) } #endif #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) # define __DEPRECATED__(msg) #elif defined(_WIN32) # define __DEPRECATED__(msg) __declspec(deprecated(msg)) #elif (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 5 && !defined(__clang__)))) # define __DEPRECATED__(msg) __attribute__((deprecated)) #else # define __DEPRECATED__(msg) __attribute__((deprecated(msg))) #endif #if defined(__CUDA_ARCH__) && !defined(__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING) # define __CDPRT_DEPRECATED(func_name) __DEPRECATED__("Use of "#func_name" from device code is deprecated. Moreover, such use will cause this module to fail to load on sm_90+ devices. If calls to "#func_name" from device code cannot be removed for older devices at this time, you may guard them with __CUDA_ARCH__ macros to remove them only for sm_90+ devices, making sure to generate code for compute_90 for the macros to take effect. Note that this mitigation will no longer work when support for "#func_name" from device code is eventually dropped for all devices. Disable this warning with -D__CDPRT_SUPPRESS_SYNC_DEPRECATION_WARNING.") #else # define __CDPRT_DEPRECATED(func_name) #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 350) #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #define cudaStreamGraphTailLaunch (cudaStream_t)0x0100000000000000 #define cudaStreamGraphFireAndForget (cudaStream_t)0x0200000000000000 #ifdef __CUDA_INTERNAL_USE_CDP2 #define cudaStreamTailLaunch ((cudaStream_t)0x3) #define cudaStreamFireAndForget ((cudaStream_t)0x4) #endif extern "C" { extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaDeviceSynchronizeDeprecationAvoidance(void); #ifndef __CUDA_INTERNAL_USE_CDP2 extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); #if (__CUDA_ARCH__ < 900) && (defined(CUDA_FORCE_CDP1_IF_SUPPORTED) || (defined(_WIN32) && !defined(_WIN64))) extern __device__ __cudart_builtin__ __CDPRT_DEPRECATED(cudaDeviceSynchronize) cudaError_t CUDARTAPI cudaDeviceSynchronize(void); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetLimit(size_t *pValue, enum cudaLimit limit); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2DeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetLastError(void); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2PeekAtLastError(void); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorString(cudaError_t error); extern __device__ __cudart_builtin__ const char* CUDARTAPI __cudaCDP2GetErrorName(cudaError_t error); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDeviceCount(int *count); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2GetDevice(int *device); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamDestroy(cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2StreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventCreateWithFlags(cudaEvent_t *event, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecord_ptsz(cudaEvent_t event, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2EventDestroy(cudaEvent_t event); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2FuncGetAttributes(struct cudaFuncAttributes *attr, const void *func); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Free(void *devPtr); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Malloc(void **devPtr, size_t size); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2MemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2Memset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2RuntimeGetVersion(int *runtimeVersion); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBuffer(size_t alignment, size_t size); extern __device__ __cudart_builtin__ void * CUDARTAPI __cudaCDP2GetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2LaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch(cudaGraphExec_t graphExec, cudaStream_t stream); #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) static inline __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGraphLaunch_ptsz(cudaGraphExec_t graphExec, cudaStream_t stream) { if (stream == 0) { stream = cudaStreamPerThread; } return cudaGraphLaunch(graphExec, stream); } #endif static inline __device__ __cudart_builtin__ cudaGraphExec_t CUDARTAPI cudaGetCurrentGraphExec(void) { unsigned long long current_graph_exec; asm ("mov.u64 %0, %%current_graph_exec;" : "=l"(current_graph_exec)); return (cudaGraphExec_t)current_graph_exec; } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaTriggerProgrammaticLaunchCompletion(void) { asm volatile("griddepcontrol.launch_dependents;":::); } static inline __device__ __cudart_builtin__ void CUDARTAPI cudaGridDependencySynchronize(void) { asm volatile("griddepcontrol.wait;":::"memory"); } extern __device__ __cudart_builtin__ unsigned long long CUDARTAPI cudaCGGetIntrinsicHandle(enum cudaCGScope scope); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronize(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGSynchronizeGrid(unsigned long long handle, unsigned int flags); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetSize(unsigned int *numThreads, unsigned int *numGrids, unsigned long long handle); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaCGGetRank(unsigned int *threadRank, unsigned int *gridRank, unsigned long long handle); #ifdef __CUDA_ARCH__ #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) { return __cudaCDP2DeviceGetAttribute(value, attr, device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, enum cudaLimit limit) { return __cudaCDP2DeviceGetLimit(pValue, limit); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetCacheConfig(enum cudaFuncCache *pCacheConfig) { return __cudaCDP2DeviceGetCacheConfig(pCacheConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaDeviceGetSharedMemConfig(enum cudaSharedMemConfig *pConfig) { return __cudaCDP2DeviceGetSharedMemConfig(pConfig); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetLastError(void) { return __cudaCDP2GetLastError(); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaPeekAtLastError(void) { return __cudaCDP2PeekAtLastError(); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) { return __cudaCDP2GetErrorString(error); } static __inline__ __device__ __cudart_builtin__ const char* CUDARTAPI cudaGetErrorName(cudaError_t error) { return __cudaCDP2GetErrorName(error); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { return __cudaCDP2GetDeviceCount(count); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { return __cudaCDP2GetDevice(device); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *pStream, unsigned int flags) { return __cudaCDP2StreamCreateWithFlags(pStream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { return __cudaCDP2StreamDestroy(stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaStreamWaitEvent_ptsz(cudaStream_t stream, cudaEvent_t event, unsigned int flags) { return __cudaCDP2StreamWaitEvent_ptsz(stream, event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventCreateWithFlags(cudaEvent_t *event, unsigned int flags) { return __cudaCDP2EventCreateWithFlags(event, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecord_ptsz(cudaEvent_t event, cudaStream_t stream) { return __cudaCDP2EventRecord_ptsz(event, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventRecordWithFlags_ptsz(cudaEvent_t event, cudaStream_t stream, unsigned int flags) { return __cudaCDP2EventRecordWithFlags_ptsz(event, stream, flags); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { return __cudaCDP2EventDestroy(event); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const void *func) { return __cudaCDP2FuncGetAttributes(attr, func); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { return __cudaCDP2Free(devPtr); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { return __cudaCDP2Malloc(devPtr, size); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpyAsync_ptsz(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2MemcpyAsync_ptsz(dst, src, count, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy2DAsync_ptsz(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) { return __cudaCDP2Memcpy2DAsync_ptsz(dst, dpitch, src, spitch, width, height, kind, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemcpy3DAsync_ptsz(const struct cudaMemcpy3DParms *p, cudaStream_t stream) { return __cudaCDP2Memcpy3DAsync_ptsz(p, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemsetAsync_ptsz(void *devPtr, int value, size_t count, cudaStream_t stream) { return __cudaCDP2MemsetAsync_ptsz(devPtr, value, count, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset2DAsync_ptsz(void *devPtr, size_t pitch, int value, size_t width, size_t height, cudaStream_t stream) { return __cudaCDP2Memset2DAsync_ptsz(devPtr, pitch, value, width, height, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaMemset3DAsync_ptsz(struct cudaPitchedPtr pitchedDevPtr, int value, struct cudaExtent extent, cudaStream_t stream) { return __cudaCDP2Memset3DAsync_ptsz(pitchedDevPtr, value, extent, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { return __cudaCDP2RuntimeGetVersion(runtimeVersion); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSmemSize); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, const void *func, int blockSize, size_t dynamicSmemSize, unsigned int flags) { return __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func, blockSize, dynamicSmemSize, flags); } #endif #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size) { return __cudaCDP2GetParameterBuffer(alignment, size); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBuffer(size_t alignment, size_t size); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize) { return __cudaCDP2GetParameterBufferV2(func, gridDimension, blockDimension, sharedMemSize); } #else extern __device__ __cudart_builtin__ void * CUDARTAPI cudaGetParameterBufferV2(void *func, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize); #endif #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice_ptsz(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2_ptsz(void *parameterBuffer, cudaStream_t stream); #endif #if defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) && defined(__CUDA_ARCH__) static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #else return cudaLaunchDevice_ptsz(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); #endif } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { #ifdef __CUDA_INTERNAL_USE_CDP2 return __cudaCDP2LaunchDeviceV2_ptsz(parameterBuffer, stream); #else return cudaLaunchDeviceV2_ptsz(parameterBuffer, stream); #endif } #else #ifdef __CUDA_INTERNAL_USE_CDP2 static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream) { return __cudaCDP2LaunchDevice(func, parameterBuffer, gridDimension, blockDimension, sharedMemSize, stream); } static __inline__ __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream) { return __cudaCDP2LaunchDeviceV2(parameterBuffer, stream); } #else extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDevice(void *func, void *parameterBuffer, dim3 gridDimension, dim3 blockDimension, unsigned int sharedMemSize, cudaStream_t stream); extern __device__ __cudart_builtin__ cudaError_t CUDARTAPI cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream); #endif #endif #define __cudaCDP2DeviceGetAttribute #define __cudaCDP2DeviceGetLimit #define __cudaCDP2DeviceGetCacheConfig #define __cudaCDP2DeviceGetSharedMemConfig #define __cudaCDP2GetLastError #define __cudaCDP2PeekAtLastError #define __cudaCDP2GetErrorString #define __cudaCDP2GetErrorName #define __cudaCDP2GetDeviceCount #define __cudaCDP2GetDevice #define __cudaCDP2StreamCreateWithFlags #define __cudaCDP2StreamDestroy #define __cudaCDP2StreamWaitEvent #define __cudaCDP2StreamWaitEvent_ptsz #define __cudaCDP2EventCreateWithFlags #define __cudaCDP2EventRecord #define __cudaCDP2EventRecord_ptsz #define __cudaCDP2EventRecordWithFlags #define __cudaCDP2EventRecordWithFlags_ptsz #define __cudaCDP2EventDestroy #define __cudaCDP2FuncGetAttributes #define __cudaCDP2Free #define __cudaCDP2Malloc #define __cudaCDP2MemcpyAsync #define __cudaCDP2MemcpyAsync_ptsz #define __cudaCDP2Memcpy2DAsync #define __cudaCDP2Memcpy2DAsync_ptsz #define __cudaCDP2Memcpy3DAsync #define __cudaCDP2Memcpy3DAsync_ptsz #define __cudaCDP2MemsetAsync #define __cudaCDP2MemsetAsync_ptsz #define __cudaCDP2Memset2DAsync #define __cudaCDP2Memset2DAsync_ptsz #define __cudaCDP2Memset3DAsync #define __cudaCDP2Memset3DAsync_ptsz #define __cudaCDP2RuntimeGetVersion #define __cudaCDP2GetParameterBuffer #define __cudaCDP2GetParameterBufferV2 #define __cudaCDP2LaunchDevice_ptsz #define __cudaCDP2LaunchDeviceV2_ptsz #define __cudaCDP2LaunchDevice #define __cudaCDP2LaunchDeviceV2 #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor #define __cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags } template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaMalloc(T **devPtr, size_t size); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaFuncGetAttributes(struct cudaFuncAttributes *attr, T *entry); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize); template <typename T> static __inline__ __device__ __cudart_builtin__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, T func, int blockSize, size_t dynamicSmemSize, unsigned int flags); #endif #endif #undef __DEPRECATED__ #undef __CDPRT_DEPRECATED #undef __CUDA_INTERNAL_USE_CDP2 #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/math_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/math_functions.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_MATH_FUNCTIONS_H__ #endif #if !defined(__MATH_FUNCTIONS_H__) #define __MATH_FUNCTIONS_H__ #if defined(__QNX__) && (__GNUC__ >= 5) && defined(__CUDACC__) #if __has_include(<__config>) #include <__config> #endif #endif #if defined(__cplusplus) && defined(__CUDACC__) #if !defined(__DEVICE_TYPES_H__) #define __DEVICE_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif enum __device_builtin__ cudaRoundMode { cudaRoundNearest, cudaRoundZero, cudaRoundPosInf, cudaRoundMinInf }; #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DEVICE_TYPES_H__ #endif #endif #if !defined(__CUDACC_RTC__) #endif #if !defined(__SURFACE_TYPES_H__) #define __SURFACE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaSurfaceType1D 0x01 #define cudaSurfaceType2D 0x02 #define cudaSurfaceType3D 0x03 #define cudaSurfaceTypeCubemap 0x0C #define cudaSurfaceType1DLayered 0xF1 #define cudaSurfaceType2DLayered 0xF2 #define cudaSurfaceTypeCubemapLayered 0xFC enum __device_builtin__ cudaSurfaceBoundaryMode { cudaBoundaryModeZero = 0, cudaBoundaryModeClamp = 1, cudaBoundaryModeTrap = 2 }; enum __device_builtin__ cudaSurfaceFormatMode { cudaFormatModeForced = 0, cudaFormatModeAuto = 1 }; typedef __device_builtin__ unsigned long long cudaSurfaceObject_t; #endif #if !defined(__TEXTURE_TYPES_H__) #define __TEXTURE_TYPES_H__ #if !defined(__DRIVER_TYPES_H__) #define __DRIVER_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INTERNAL_COMPILATION__) #if !defined(__CUDACC_RTC__) #include <limits.h> #include <stddef.h> #endif #define cudaHostAllocDefault 0x00 #define cudaHostAllocPortable 0x01 #define cudaHostAllocMapped 0x02 #define cudaHostAllocWriteCombined 0x04 #define cudaHostRegisterDefault 0x00 #define cudaHostRegisterPortable 0x01 #define cudaHostRegisterMapped 0x02 #define cudaHostRegisterIoMemory 0x04 #define cudaHostRegisterReadOnly 0x08 #define cudaPeerAccessDefault 0x00 #define cudaStreamDefault 0x00 #define cudaStreamNonBlocking 0x01 #define cudaStreamLegacy ((cudaStream_t)0x1) #define cudaStreamPerThread ((cudaStream_t)0x2) #define cudaEventDefault 0x00 #define cudaEventBlockingSync 0x01 #define cudaEventDisableTiming 0x02 #define cudaEventInterprocess 0x04 #define cudaEventRecordDefault 0x00 #define cudaEventRecordExternal 0x01 #define cudaEventWaitDefault 0x00 #define cudaEventWaitExternal 0x01 #define cudaDeviceScheduleAuto 0x00 #define cudaDeviceScheduleSpin 0x01 #define cudaDeviceScheduleYield 0x02 #define cudaDeviceScheduleBlockingSync 0x04 #define cudaDeviceBlockingSync 0x04 #define cudaDeviceScheduleMask 0x07 #define cudaDeviceMapHost 0x08 #define cudaDeviceLmemResizeToMax 0x10 #define cudaDeviceSyncMemops 0x80 #define cudaDeviceMask 0xff #define cudaArrayDefault 0x00 #define cudaArrayLayered 0x01 #define cudaArraySurfaceLoadStore 0x02 #define cudaArrayCubemap 0x04 #define cudaArrayTextureGather 0x08 #define cudaArrayColorAttachment 0x20 #define cudaArraySparse 0x40 #define cudaArrayDeferredMapping 0x80 #define cudaIpcMemLazyEnablePeerAccess 0x01 #define cudaMemAttachGlobal 0x01 #define cudaMemAttachHost 0x02 #define cudaMemAttachSingle 0x04 #define cudaOccupancyDefault 0x00 #define cudaOccupancyDisableCachingOverride 0x01 #define cudaCpuDeviceId ((int)-1) #define cudaInvalidDeviceId ((int)-2) #define cudaInitDeviceFlagsAreValid 0x01 #define cudaCooperativeLaunchMultiDeviceNoPreSync 0x01 #define cudaCooperativeLaunchMultiDeviceNoPostSync 0x02 #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif enum __device_builtin__ cudaError { cudaSuccess = 0, cudaErrorInvalidValue = 1, cudaErrorMemoryAllocation = 2, cudaErrorInitializationError = 3, cudaErrorCudartUnloading = 4, cudaErrorProfilerDisabled = 5, cudaErrorProfilerNotInitialized = 6, cudaErrorProfilerAlreadyStarted = 7, cudaErrorProfilerAlreadyStopped = 8, cudaErrorInvalidConfiguration = 9, cudaErrorInvalidPitchValue = 12, cudaErrorInvalidSymbol = 13, cudaErrorInvalidHostPointer = 16, cudaErrorInvalidDevicePointer = 17, cudaErrorInvalidTexture = 18, cudaErrorInvalidTextureBinding = 19, cudaErrorInvalidChannelDescriptor = 20, cudaErrorInvalidMemcpyDirection = 21, cudaErrorAddressOfConstant = 22, cudaErrorTextureFetchFailed = 23, cudaErrorTextureNotBound = 24, cudaErrorSynchronizationError = 25, cudaErrorInvalidFilterSetting = 26, cudaErrorInvalidNormSetting = 27, cudaErrorMixedDeviceExecution = 28, cudaErrorNotYetImplemented = 31, cudaErrorMemoryValueTooLarge = 32, cudaErrorStubLibrary = 34, cudaErrorInsufficientDriver = 35, cudaErrorCallRequiresNewerDriver = 36, cudaErrorInvalidSurface = 37, cudaErrorDuplicateVariableName = 43, cudaErrorDuplicateTextureName = 44, cudaErrorDuplicateSurfaceName = 45, cudaErrorDevicesUnavailable = 46, cudaErrorIncompatibleDriverContext = 49, cudaErrorMissingConfiguration = 52, cudaErrorPriorLaunchFailure = 53, cudaErrorLaunchMaxDepthExceeded = 65, cudaErrorLaunchFileScopedTex = 66, cudaErrorLaunchFileScopedSurf = 67, cudaErrorSyncDepthExceeded = 68, cudaErrorLaunchPendingCountExceeded = 69, cudaErrorInvalidDeviceFunction = 98, cudaErrorNoDevice = 100, cudaErrorInvalidDevice = 101, cudaErrorDeviceNotLicensed = 102, cudaErrorSoftwareValidityNotEstablished = 103, cudaErrorStartupFailure = 127, cudaErrorInvalidKernelImage = 200, cudaErrorDeviceUninitialized = 201, cudaErrorMapBufferObjectFailed = 205, cudaErrorUnmapBufferObjectFailed = 206, cudaErrorArrayIsMapped = 207, cudaErrorAlreadyMapped = 208, cudaErrorNoKernelImageForDevice = 209, cudaErrorAlreadyAcquired = 210, cudaErrorNotMapped = 211, cudaErrorNotMappedAsArray = 212, cudaErrorNotMappedAsPointer = 213, cudaErrorECCUncorrectable = 214, cudaErrorUnsupportedLimit = 215, cudaErrorDeviceAlreadyInUse = 216, cudaErrorPeerAccessUnsupported = 217, cudaErrorInvalidPtx = 218, cudaErrorInvalidGraphicsContext = 219, cudaErrorNvlinkUncorrectable = 220, cudaErrorJitCompilerNotFound = 221, cudaErrorUnsupportedPtxVersion = 222, cudaErrorJitCompilationDisabled = 223, cudaErrorUnsupportedExecAffinity = 224, cudaErrorUnsupportedDevSideSync = 225, cudaErrorInvalidSource = 300, cudaErrorFileNotFound = 301, cudaErrorSharedObjectSymbolNotFound = 302, cudaErrorSharedObjectInitFailed = 303, cudaErrorOperatingSystem = 304, cudaErrorInvalidResourceHandle = 400, cudaErrorIllegalState = 401, cudaErrorSymbolNotFound = 500, cudaErrorNotReady = 600, cudaErrorIllegalAddress = 700, cudaErrorLaunchOutOfResources = 701, cudaErrorLaunchTimeout = 702, cudaErrorLaunchIncompatibleTexturing = 703, cudaErrorPeerAccessAlreadyEnabled = 704, cudaErrorPeerAccessNotEnabled = 705, cudaErrorSetOnActiveProcess = 708, cudaErrorContextIsDestroyed = 709, cudaErrorAssert = 710, cudaErrorTooManyPeers = 711, cudaErrorHostMemoryAlreadyRegistered = 712, cudaErrorHostMemoryNotRegistered = 713, cudaErrorHardwareStackError = 714, cudaErrorIllegalInstruction = 715, cudaErrorMisalignedAddress = 716, cudaErrorInvalidAddressSpace = 717, cudaErrorInvalidPc = 718, cudaErrorLaunchFailure = 719, cudaErrorCooperativeLaunchTooLarge = 720, cudaErrorNotPermitted = 800, cudaErrorNotSupported = 801, cudaErrorSystemNotReady = 802, cudaErrorSystemDriverMismatch = 803, cudaErrorCompatNotSupportedOnDevice = 804, cudaErrorMpsConnectionFailed = 805, cudaErrorMpsRpcFailure = 806, cudaErrorMpsServerNotReady = 807, cudaErrorMpsMaxClientsReached = 808, cudaErrorMpsMaxConnectionsReached = 809, cudaErrorMpsClientTerminated = 810, cudaErrorCdpNotSupported = 811, cudaErrorCdpVersionMismatch = 812, cudaErrorStreamCaptureUnsupported = 900, cudaErrorStreamCaptureInvalidated = 901, cudaErrorStreamCaptureMerge = 902, cudaErrorStreamCaptureUnmatched = 903, cudaErrorStreamCaptureUnjoined = 904, cudaErrorStreamCaptureIsolation = 905, cudaErrorStreamCaptureImplicit = 906, cudaErrorCapturedEvent = 907, cudaErrorStreamCaptureWrongThread = 908, cudaErrorTimeout = 909, cudaErrorGraphExecUpdateFailure = 910, cudaErrorExternalDevice = 911, cudaErrorInvalidClusterSize = 912, cudaErrorUnknown = 999, cudaErrorApiFailureBase = 10000 }; enum __device_builtin__ cudaChannelFormatKind { cudaChannelFormatKindSigned = 0, cudaChannelFormatKindUnsigned = 1, cudaChannelFormatKindFloat = 2, cudaChannelFormatKindNone = 3, cudaChannelFormatKindNV12 = 4, cudaChannelFormatKindUnsignedNormalized8X1 = 5, cudaChannelFormatKindUnsignedNormalized8X2 = 6, cudaChannelFormatKindUnsignedNormalized8X4 = 7, cudaChannelFormatKindUnsignedNormalized16X1 = 8, cudaChannelFormatKindUnsignedNormalized16X2 = 9, cudaChannelFormatKindUnsignedNormalized16X4 = 10, cudaChannelFormatKindSignedNormalized8X1 = 11, cudaChannelFormatKindSignedNormalized8X2 = 12, cudaChannelFormatKindSignedNormalized8X4 = 13, cudaChannelFormatKindSignedNormalized16X1 = 14, cudaChannelFormatKindSignedNormalized16X2 = 15, cudaChannelFormatKindSignedNormalized16X4 = 16, cudaChannelFormatKindUnsignedBlockCompressed1 = 17, cudaChannelFormatKindUnsignedBlockCompressed1SRGB = 18, cudaChannelFormatKindUnsignedBlockCompressed2 = 19, cudaChannelFormatKindUnsignedBlockCompressed2SRGB = 20, cudaChannelFormatKindUnsignedBlockCompressed3 = 21, cudaChannelFormatKindUnsignedBlockCompressed3SRGB = 22, cudaChannelFormatKindUnsignedBlockCompressed4 = 23, cudaChannelFormatKindSignedBlockCompressed4 = 24, cudaChannelFormatKindUnsignedBlockCompressed5 = 25, cudaChannelFormatKindSignedBlockCompressed5 = 26, cudaChannelFormatKindUnsignedBlockCompressed6H = 27, cudaChannelFormatKindSignedBlockCompressed6H = 28, cudaChannelFormatKindUnsignedBlockCompressed7 = 29, cudaChannelFormatKindUnsignedBlockCompressed7SRGB = 30 }; struct __device_builtin__ cudaChannelFormatDesc { int x; int y; int z; int w; enum cudaChannelFormatKind f; }; typedef struct cudaArray *cudaArray_t; typedef const struct cudaArray *cudaArray_const_t; struct cudaArray; typedef struct cudaMipmappedArray *cudaMipmappedArray_t; typedef const struct cudaMipmappedArray *cudaMipmappedArray_const_t; struct cudaMipmappedArray; #define cudaArraySparsePropertiesSingleMipTail 0x1 struct __device_builtin__ cudaArraySparseProperties { struct { unsigned int width; unsigned int height; unsigned int depth; } tileExtent; unsigned int miptailFirstLevel; unsigned long long miptailSize; unsigned int flags; unsigned int reserved[4]; }; struct __device_builtin__ cudaArrayMemoryRequirements { size_t size; size_t alignment; unsigned int reserved[4]; }; enum __device_builtin__ cudaMemoryType { cudaMemoryTypeUnregistered = 0, cudaMemoryTypeHost = 1, cudaMemoryTypeDevice = 2, cudaMemoryTypeManaged = 3 }; enum __device_builtin__ cudaMemcpyKind { cudaMemcpyHostToHost = 0, cudaMemcpyHostToDevice = 1, cudaMemcpyDeviceToHost = 2, cudaMemcpyDeviceToDevice = 3, cudaMemcpyDefault = 4 }; struct __device_builtin__ cudaPitchedPtr { void *ptr; size_t pitch; size_t xsize; size_t ysize; }; struct __device_builtin__ cudaExtent { size_t width; size_t height; size_t depth; }; struct __device_builtin__ cudaPos { size_t x; size_t y; size_t z; }; struct __device_builtin__ cudaMemcpy3DParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; struct cudaExtent extent; enum cudaMemcpyKind kind; }; struct __device_builtin__ cudaMemcpy3DPeerParms { cudaArray_t srcArray; struct cudaPos srcPos; struct cudaPitchedPtr srcPtr; int srcDevice; cudaArray_t dstArray; struct cudaPos dstPos; struct cudaPitchedPtr dstPtr; int dstDevice; struct cudaExtent extent; }; struct __device_builtin__ cudaMemsetParams { void *dst; size_t pitch; unsigned int value; unsigned int elementSize; size_t width; size_t height; }; enum __device_builtin__ cudaAccessProperty { cudaAccessPropertyNormal = 0, cudaAccessPropertyStreaming = 1, cudaAccessPropertyPersisting = 2 }; struct __device_builtin__ cudaAccessPolicyWindow { void *base_ptr; size_t num_bytes; float hitRatio; enum cudaAccessProperty hitProp; enum cudaAccessProperty missProp; }; #ifdef _WIN32 #define CUDART_CB __stdcall #else #define CUDART_CB #endif typedef void (CUDART_CB *cudaHostFn_t)(void *userData); struct __device_builtin__ cudaHostNodeParams { cudaHostFn_t fn; void* userData; }; enum __device_builtin__ cudaStreamCaptureStatus { cudaStreamCaptureStatusNone = 0, cudaStreamCaptureStatusActive = 1, cudaStreamCaptureStatusInvalidated = 2 }; enum __device_builtin__ cudaStreamCaptureMode { cudaStreamCaptureModeGlobal = 0, cudaStreamCaptureModeThreadLocal = 1, cudaStreamCaptureModeRelaxed = 2 }; enum __device_builtin__ cudaSynchronizationPolicy { cudaSyncPolicyAuto = 1, cudaSyncPolicySpin = 2, cudaSyncPolicyYield = 3, cudaSyncPolicyBlockingSync = 4 }; enum __device_builtin__ cudaClusterSchedulingPolicy { cudaClusterSchedulingPolicyDefault = 0, cudaClusterSchedulingPolicySpread = 1, cudaClusterSchedulingPolicyLoadBalancing = 2 }; enum __device_builtin__ cudaStreamUpdateCaptureDependenciesFlags { cudaStreamAddCaptureDependencies = 0x0, cudaStreamSetCaptureDependencies = 0x1 }; enum __device_builtin__ cudaUserObjectFlags { cudaUserObjectNoDestructorSync = 0x1 }; enum __device_builtin__ cudaUserObjectRetainFlags { cudaGraphUserObjectMove = 0x1 }; struct cudaGraphicsResource; enum __device_builtin__ cudaGraphicsRegisterFlags { cudaGraphicsRegisterFlagsNone = 0, cudaGraphicsRegisterFlagsReadOnly = 1, cudaGraphicsRegisterFlagsWriteDiscard = 2, cudaGraphicsRegisterFlagsSurfaceLoadStore = 4, cudaGraphicsRegisterFlagsTextureGather = 8 }; enum __device_builtin__ cudaGraphicsMapFlags { cudaGraphicsMapFlagsNone = 0, cudaGraphicsMapFlagsReadOnly = 1, cudaGraphicsMapFlagsWriteDiscard = 2 }; enum __device_builtin__ cudaGraphicsCubeFace { cudaGraphicsCubeFacePositiveX = 0x00, cudaGraphicsCubeFaceNegativeX = 0x01, cudaGraphicsCubeFacePositiveY = 0x02, cudaGraphicsCubeFaceNegativeY = 0x03, cudaGraphicsCubeFacePositiveZ = 0x04, cudaGraphicsCubeFaceNegativeZ = 0x05 }; enum __device_builtin__ cudaResourceType { cudaResourceTypeArray = 0x00, cudaResourceTypeMipmappedArray = 0x01, cudaResourceTypeLinear = 0x02, cudaResourceTypePitch2D = 0x03 }; enum __device_builtin__ cudaResourceViewFormat { cudaResViewFormatNone = 0x00, cudaResViewFormatUnsignedChar1 = 0x01, cudaResViewFormatUnsignedChar2 = 0x02, cudaResViewFormatUnsignedChar4 = 0x03, cudaResViewFormatSignedChar1 = 0x04, cudaResViewFormatSignedChar2 = 0x05, cudaResViewFormatSignedChar4 = 0x06, cudaResViewFormatUnsignedShort1 = 0x07, cudaResViewFormatUnsignedShort2 = 0x08, cudaResViewFormatUnsignedShort4 = 0x09, cudaResViewFormatSignedShort1 = 0x0a, cudaResViewFormatSignedShort2 = 0x0b, cudaResViewFormatSignedShort4 = 0x0c, cudaResViewFormatUnsignedInt1 = 0x0d, cudaResViewFormatUnsignedInt2 = 0x0e, cudaResViewFormatUnsignedInt4 = 0x0f, cudaResViewFormatSignedInt1 = 0x10, cudaResViewFormatSignedInt2 = 0x11, cudaResViewFormatSignedInt4 = 0x12, cudaResViewFormatHalf1 = 0x13, cudaResViewFormatHalf2 = 0x14, cudaResViewFormatHalf4 = 0x15, cudaResViewFormatFloat1 = 0x16, cudaResViewFormatFloat2 = 0x17, cudaResViewFormatFloat4 = 0x18, cudaResViewFormatUnsignedBlockCompressed1 = 0x19, cudaResViewFormatUnsignedBlockCompressed2 = 0x1a, cudaResViewFormatUnsignedBlockCompressed3 = 0x1b, cudaResViewFormatUnsignedBlockCompressed4 = 0x1c, cudaResViewFormatSignedBlockCompressed4 = 0x1d, cudaResViewFormatUnsignedBlockCompressed5 = 0x1e, cudaResViewFormatSignedBlockCompressed5 = 0x1f, cudaResViewFormatUnsignedBlockCompressed6H = 0x20, cudaResViewFormatSignedBlockCompressed6H = 0x21, cudaResViewFormatUnsignedBlockCompressed7 = 0x22 }; struct __device_builtin__ cudaResourceDesc { enum cudaResourceType resType; union { struct { cudaArray_t array; } array; struct { cudaMipmappedArray_t mipmap; } mipmap; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t sizeInBytes; } linear; struct { void *devPtr; struct cudaChannelFormatDesc desc; size_t width; size_t height; size_t pitchInBytes; } pitch2D; } res; }; struct __device_builtin__ cudaResourceViewDesc { enum cudaResourceViewFormat format; size_t width; size_t height; size_t depth; unsigned int firstMipmapLevel; unsigned int lastMipmapLevel; unsigned int firstLayer; unsigned int lastLayer; }; struct __device_builtin__ cudaPointerAttributes { enum cudaMemoryType type; int device; void *devicePointer; void *hostPointer; }; struct __device_builtin__ cudaFuncAttributes { size_t sharedSizeBytes; size_t constSizeBytes; size_t localSizeBytes; int maxThreadsPerBlock; int numRegs; int ptxVersion; int binaryVersion; int cacheModeCA; int maxDynamicSharedSizeBytes; int preferredShmemCarveout; int clusterDimMustBeSet; int requiredClusterWidth; int requiredClusterHeight; int requiredClusterDepth; int clusterSchedulingPolicyPreference; int nonPortableClusterSizeAllowed; int reserved[16]; }; enum __device_builtin__ cudaFuncAttribute { cudaFuncAttributeMaxDynamicSharedMemorySize = 8, cudaFuncAttributePreferredSharedMemoryCarveout = 9, cudaFuncAttributeClusterDimMustBeSet = 10, cudaFuncAttributeRequiredClusterWidth = 11, cudaFuncAttributeRequiredClusterHeight = 12, cudaFuncAttributeRequiredClusterDepth = 13, cudaFuncAttributeNonPortableClusterSizeAllowed = 14, cudaFuncAttributeClusterSchedulingPolicyPreference = 15, cudaFuncAttributeMax }; enum __device_builtin__ cudaFuncCache { cudaFuncCachePreferNone = 0, cudaFuncCachePreferShared = 1, cudaFuncCachePreferL1 = 2, cudaFuncCachePreferEqual = 3 }; enum __device_builtin__ cudaSharedMemConfig { cudaSharedMemBankSizeDefault = 0, cudaSharedMemBankSizeFourByte = 1, cudaSharedMemBankSizeEightByte = 2 }; enum __device_builtin__ cudaSharedCarveout { cudaSharedmemCarveoutDefault = -1, cudaSharedmemCarveoutMaxShared = 100, cudaSharedmemCarveoutMaxL1 = 0 }; enum __device_builtin__ cudaComputeMode { cudaComputeModeDefault = 0, cudaComputeModeExclusive = 1, cudaComputeModeProhibited = 2, cudaComputeModeExclusiveProcess = 3 }; enum __device_builtin__ cudaLimit { cudaLimitStackSize = 0x00, cudaLimitPrintfFifoSize = 0x01, cudaLimitMallocHeapSize = 0x02, cudaLimitDevRuntimeSyncDepth = 0x03, cudaLimitDevRuntimePendingLaunchCount = 0x04, cudaLimitMaxL2FetchGranularity = 0x05, cudaLimitPersistingL2CacheSize = 0x06 }; enum __device_builtin__ cudaMemoryAdvise { cudaMemAdviseSetReadMostly = 1, cudaMemAdviseUnsetReadMostly = 2, cudaMemAdviseSetPreferredLocation = 3, cudaMemAdviseUnsetPreferredLocation = 4, cudaMemAdviseSetAccessedBy = 5, cudaMemAdviseUnsetAccessedBy = 6 }; enum __device_builtin__ cudaMemRangeAttribute { cudaMemRangeAttributeReadMostly = 1, cudaMemRangeAttributePreferredLocation = 2, cudaMemRangeAttributeAccessedBy = 3, cudaMemRangeAttributeLastPrefetchLocation = 4 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesOptions { cudaFlushGPUDirectRDMAWritesOptionHost = 1<<0, cudaFlushGPUDirectRDMAWritesOptionMemOps = 1<<1 }; enum __device_builtin__ cudaGPUDirectRDMAWritesOrdering { cudaGPUDirectRDMAWritesOrderingNone = 0, cudaGPUDirectRDMAWritesOrderingOwner = 100, cudaGPUDirectRDMAWritesOrderingAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesScope { cudaFlushGPUDirectRDMAWritesToOwner = 100, cudaFlushGPUDirectRDMAWritesToAllDevices = 200 }; enum __device_builtin__ cudaFlushGPUDirectRDMAWritesTarget { cudaFlushGPUDirectRDMAWritesTargetCurrentDevice }; enum __device_builtin__ cudaDeviceAttr { cudaDevAttrMaxThreadsPerBlock = 1, cudaDevAttrMaxBlockDimX = 2, cudaDevAttrMaxBlockDimY = 3, cudaDevAttrMaxBlockDimZ = 4, cudaDevAttrMaxGridDimX = 5, cudaDevAttrMaxGridDimY = 6, cudaDevAttrMaxGridDimZ = 7, cudaDevAttrMaxSharedMemoryPerBlock = 8, cudaDevAttrTotalConstantMemory = 9, cudaDevAttrWarpSize = 10, cudaDevAttrMaxPitch = 11, cudaDevAttrMaxRegistersPerBlock = 12, cudaDevAttrClockRate = 13, cudaDevAttrTextureAlignment = 14, cudaDevAttrGpuOverlap = 15, cudaDevAttrMultiProcessorCount = 16, cudaDevAttrKernelExecTimeout = 17, cudaDevAttrIntegrated = 18, cudaDevAttrCanMapHostMemory = 19, cudaDevAttrComputeMode = 20, cudaDevAttrMaxTexture1DWidth = 21, cudaDevAttrMaxTexture2DWidth = 22, cudaDevAttrMaxTexture2DHeight = 23, cudaDevAttrMaxTexture3DWidth = 24, cudaDevAttrMaxTexture3DHeight = 25, cudaDevAttrMaxTexture3DDepth = 26, cudaDevAttrMaxTexture2DLayeredWidth = 27, cudaDevAttrMaxTexture2DLayeredHeight = 28, cudaDevAttrMaxTexture2DLayeredLayers = 29, cudaDevAttrSurfaceAlignment = 30, cudaDevAttrConcurrentKernels = 31, cudaDevAttrEccEnabled = 32, cudaDevAttrPciBusId = 33, cudaDevAttrPciDeviceId = 34, cudaDevAttrTccDriver = 35, cudaDevAttrMemoryClockRate = 36, cudaDevAttrGlobalMemoryBusWidth = 37, cudaDevAttrL2CacheSize = 38, cudaDevAttrMaxThreadsPerMultiProcessor = 39, cudaDevAttrAsyncEngineCount = 40, cudaDevAttrUnifiedAddressing = 41, cudaDevAttrMaxTexture1DLayeredWidth = 42, cudaDevAttrMaxTexture1DLayeredLayers = 43, cudaDevAttrMaxTexture2DGatherWidth = 45, cudaDevAttrMaxTexture2DGatherHeight = 46, cudaDevAttrMaxTexture3DWidthAlt = 47, cudaDevAttrMaxTexture3DHeightAlt = 48, cudaDevAttrMaxTexture3DDepthAlt = 49, cudaDevAttrPciDomainId = 50, cudaDevAttrTexturePitchAlignment = 51, cudaDevAttrMaxTextureCubemapWidth = 52, cudaDevAttrMaxTextureCubemapLayeredWidth = 53, cudaDevAttrMaxTextureCubemapLayeredLayers = 54, cudaDevAttrMaxSurface1DWidth = 55, cudaDevAttrMaxSurface2DWidth = 56, cudaDevAttrMaxSurface2DHeight = 57, cudaDevAttrMaxSurface3DWidth = 58, cudaDevAttrMaxSurface3DHeight = 59, cudaDevAttrMaxSurface3DDepth = 60, cudaDevAttrMaxSurface1DLayeredWidth = 61, cudaDevAttrMaxSurface1DLayeredLayers = 62, cudaDevAttrMaxSurface2DLayeredWidth = 63, cudaDevAttrMaxSurface2DLayeredHeight = 64, cudaDevAttrMaxSurface2DLayeredLayers = 65, cudaDevAttrMaxSurfaceCubemapWidth = 66, cudaDevAttrMaxSurfaceCubemapLayeredWidth = 67, cudaDevAttrMaxSurfaceCubemapLayeredLayers = 68, cudaDevAttrMaxTexture1DLinearWidth = 69, cudaDevAttrMaxTexture2DLinearWidth = 70, cudaDevAttrMaxTexture2DLinearHeight = 71, cudaDevAttrMaxTexture2DLinearPitch = 72, cudaDevAttrMaxTexture2DMipmappedWidth = 73, cudaDevAttrMaxTexture2DMipmappedHeight = 74, cudaDevAttrComputeCapabilityMajor = 75, cudaDevAttrComputeCapabilityMinor = 76, cudaDevAttrMaxTexture1DMipmappedWidth = 77, cudaDevAttrStreamPrioritiesSupported = 78, cudaDevAttrGlobalL1CacheSupported = 79, cudaDevAttrLocalL1CacheSupported = 80, cudaDevAttrMaxSharedMemoryPerMultiprocessor = 81, cudaDevAttrMaxRegistersPerMultiprocessor = 82, cudaDevAttrManagedMemory = 83, cudaDevAttrIsMultiGpuBoard = 84, cudaDevAttrMultiGpuBoardGroupID = 85, cudaDevAttrHostNativeAtomicSupported = 86, cudaDevAttrSingleToDoublePrecisionPerfRatio = 87, cudaDevAttrPageableMemoryAccess = 88, cudaDevAttrConcurrentManagedAccess = 89, cudaDevAttrComputePreemptionSupported = 90, cudaDevAttrCanUseHostPointerForRegisteredMem = 91, cudaDevAttrReserved92 = 92, cudaDevAttrReserved93 = 93, cudaDevAttrReserved94 = 94, cudaDevAttrCooperativeLaunch = 95, cudaDevAttrCooperativeMultiDeviceLaunch = 96, cudaDevAttrMaxSharedMemoryPerBlockOptin = 97, cudaDevAttrCanFlushRemoteWrites = 98, cudaDevAttrHostRegisterSupported = 99, cudaDevAttrPageableMemoryAccessUsesHostPageTables = 100, cudaDevAttrDirectManagedMemAccessFromHost = 101, cudaDevAttrMaxBlocksPerMultiprocessor = 106, cudaDevAttrMaxPersistingL2CacheSize = 108, cudaDevAttrMaxAccessPolicyWindowSize = 109, cudaDevAttrReservedSharedMemoryPerBlock = 111, cudaDevAttrSparseCudaArraySupported = 112, cudaDevAttrHostRegisterReadOnlySupported = 113, cudaDevAttrTimelineSemaphoreInteropSupported = 114, cudaDevAttrMaxTimelineSemaphoreInteropSupported = 114, cudaDevAttrMemoryPoolsSupported = 115, cudaDevAttrGPUDirectRDMASupported = 116, cudaDevAttrGPUDirectRDMAWritesOrdering = 118, cudaDevAttrMemoryPoolSupportedHandleTypes = 119, cudaDevAttrClusterLaunch = 120, cudaDevAttrDeferredMappingCudaArraySupported = 121, cudaDevAttrReserved122 = 122, cudaDevAttrReserved123 = 123, cudaDevAttrReserved124 = 124, cudaDevAttrIpcEventSupport = 125, cudaDevAttrMemSyncDomainCount = 126, cudaDevAttrReserved127 = 127, cudaDevAttrReserved128 = 128, cudaDevAttrReserved129 = 129, cudaDevAttrReserved132 = 132, cudaDevAttrMax }; enum __device_builtin__ cudaMemPoolAttr { cudaMemPoolReuseFollowEventDependencies = 0x1, cudaMemPoolReuseAllowOpportunistic = 0x2, cudaMemPoolReuseAllowInternalDependencies = 0x3, cudaMemPoolAttrReleaseThreshold = 0x4, cudaMemPoolAttrReservedMemCurrent = 0x5, cudaMemPoolAttrReservedMemHigh = 0x6, cudaMemPoolAttrUsedMemCurrent = 0x7, cudaMemPoolAttrUsedMemHigh = 0x8 }; enum __device_builtin__ cudaMemLocationType { cudaMemLocationTypeInvalid = 0, cudaMemLocationTypeDevice = 1 }; struct __device_builtin__ cudaMemLocation { enum cudaMemLocationType type; int id; }; enum __device_builtin__ cudaMemAccessFlags { cudaMemAccessFlagsProtNone = 0, cudaMemAccessFlagsProtRead = 1, cudaMemAccessFlagsProtReadWrite = 3 }; struct __device_builtin__ cudaMemAccessDesc { struct cudaMemLocation location; enum cudaMemAccessFlags flags; }; enum __device_builtin__ cudaMemAllocationType { cudaMemAllocationTypeInvalid = 0x0, cudaMemAllocationTypePinned = 0x1, cudaMemAllocationTypeMax = 0x7FFFFFFF }; enum __device_builtin__ cudaMemAllocationHandleType { cudaMemHandleTypeNone = 0x0, cudaMemHandleTypePosixFileDescriptor = 0x1, cudaMemHandleTypeWin32 = 0x2, cudaMemHandleTypeWin32Kmt = 0x4 }; struct __device_builtin__ cudaMemPoolProps { enum cudaMemAllocationType allocType; enum cudaMemAllocationHandleType handleTypes; struct cudaMemLocation location; void *win32SecurityAttributes; unsigned char reserved[64]; }; struct __device_builtin__ cudaMemPoolPtrExportData { unsigned char reserved[64]; }; struct __device_builtin__ cudaMemAllocNodeParams { struct cudaMemPoolProps poolProps; const struct cudaMemAccessDesc *accessDescs; size_t accessDescCount; size_t bytesize; void *dptr; }; enum __device_builtin__ cudaGraphMemAttributeType { cudaGraphMemAttrUsedMemCurrent = 0x0, cudaGraphMemAttrUsedMemHigh = 0x1, cudaGraphMemAttrReservedMemCurrent = 0x2, cudaGraphMemAttrReservedMemHigh = 0x3 }; enum __device_builtin__ cudaDeviceP2PAttr { cudaDevP2PAttrPerformanceRank = 1, cudaDevP2PAttrAccessSupported = 2, cudaDevP2PAttrNativeAtomicSupported = 3, cudaDevP2PAttrCudaArrayAccessSupported = 4 }; #ifndef CU_UUID_HAS_BEEN_DEFINED #define CU_UUID_HAS_BEEN_DEFINED struct __device_builtin__ CUuuid_st { char bytes[16]; }; typedef __device_builtin__ struct CUuuid_st CUuuid; #endif typedef __device_builtin__ struct CUuuid_st cudaUUID_t; struct __device_builtin__ cudaDeviceProp { char name[256]; cudaUUID_t uuid; char luid[8]; unsigned int luidDeviceNodeMask; size_t totalGlobalMem; size_t sharedMemPerBlock; int regsPerBlock; int warpSize; size_t memPitch; int maxThreadsPerBlock; int maxThreadsDim[3]; int maxGridSize[3]; int clockRate; size_t totalConstMem; int major; int minor; size_t textureAlignment; size_t texturePitchAlignment; int deviceOverlap; int multiProcessorCount; int kernelExecTimeoutEnabled; int integrated; int canMapHostMemory; int computeMode; int maxTexture1D; int maxTexture1DMipmap; int maxTexture1DLinear; int maxTexture2D[2]; int maxTexture2DMipmap[2]; int maxTexture2DLinear[3]; int maxTexture2DGather[2]; int maxTexture3D[3]; int maxTexture3DAlt[3]; int maxTextureCubemap; int maxTexture1DLayered[2]; int maxTexture2DLayered[3]; int maxTextureCubemapLayered[2]; int maxSurface1D; int maxSurface2D[2]; int maxSurface3D[3]; int maxSurface1DLayered[2]; int maxSurface2DLayered[3]; int maxSurfaceCubemap; int maxSurfaceCubemapLayered[2]; size_t surfaceAlignment; int concurrentKernels; int ECCEnabled; int pciBusID; int pciDeviceID; int pciDomainID; int tccDriver; int asyncEngineCount; int unifiedAddressing; int memoryClockRate; int memoryBusWidth; int l2CacheSize; int persistingL2CacheMaxSize; int maxThreadsPerMultiProcessor; int streamPrioritiesSupported; int globalL1CacheSupported; int localL1CacheSupported; size_t sharedMemPerMultiprocessor; int regsPerMultiprocessor; int managedMemory; int isMultiGpuBoard; int multiGpuBoardGroupID; int hostNativeAtomicSupported; int singleToDoublePrecisionPerfRatio; int pageableMemoryAccess; int concurrentManagedAccess; int computePreemptionSupported; int canUseHostPointerForRegisteredMem; int cooperativeLaunch; int cooperativeMultiDeviceLaunch; size_t sharedMemPerBlockOptin; int pageableMemoryAccessUsesHostPageTables; int directManagedMemAccessFromHost; int maxBlocksPerMultiProcessor; int accessPolicyMaxWindowSize; size_t reservedSharedMemPerBlock; int hostRegisterSupported; int sparseCudaArraySupported; int hostRegisterReadOnlySupported; int timelineSemaphoreInteropSupported; int memoryPoolsSupported; int gpuDirectRDMASupported; unsigned int gpuDirectRDMAFlushWritesOptions; int gpuDirectRDMAWritesOrdering; unsigned int memoryPoolSupportedHandleTypes; int deferredMappingCudaArraySupported; int ipcEventSupported; int clusterLaunch; int unifiedFunctionPointers; int reserved2[2]; int reserved[61]; }; #define CUDA_IPC_HANDLE_SIZE 64 typedef __device_builtin__ struct __device_builtin__ cudaIpcEventHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcEventHandle_t; typedef __device_builtin__ struct __device_builtin__ cudaIpcMemHandle_st { char reserved[CUDA_IPC_HANDLE_SIZE]; }cudaIpcMemHandle_t; enum __device_builtin__ cudaExternalMemoryHandleType { cudaExternalMemoryHandleTypeOpaqueFd = 1, cudaExternalMemoryHandleTypeOpaqueWin32 = 2, cudaExternalMemoryHandleTypeOpaqueWin32Kmt = 3, cudaExternalMemoryHandleTypeD3D12Heap = 4, cudaExternalMemoryHandleTypeD3D12Resource = 5, cudaExternalMemoryHandleTypeD3D11Resource = 6, cudaExternalMemoryHandleTypeD3D11ResourceKmt = 7, cudaExternalMemoryHandleTypeNvSciBuf = 8 }; #define cudaExternalMemoryDedicated 0x1 #define cudaExternalSemaphoreSignalSkipNvSciBufMemSync 0x01 #define cudaExternalSemaphoreWaitSkipNvSciBufMemSync 0x02 #define cudaNvSciSyncAttrSignal 0x1 #define cudaNvSciSyncAttrWait 0x2 struct __device_builtin__ cudaExternalMemoryHandleDesc { enum cudaExternalMemoryHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void *nvSciBufObject; } handle; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryBufferDesc { unsigned long long offset; unsigned long long size; unsigned int flags; }; struct __device_builtin__ cudaExternalMemoryMipmappedArrayDesc { unsigned long long offset; struct cudaChannelFormatDesc formatDesc; struct cudaExtent extent; unsigned int flags; unsigned int numLevels; }; enum __device_builtin__ cudaExternalSemaphoreHandleType { cudaExternalSemaphoreHandleTypeOpaqueFd = 1, cudaExternalSemaphoreHandleTypeOpaqueWin32 = 2, cudaExternalSemaphoreHandleTypeOpaqueWin32Kmt = 3, cudaExternalSemaphoreHandleTypeD3D12Fence = 4, cudaExternalSemaphoreHandleTypeD3D11Fence = 5, cudaExternalSemaphoreHandleTypeNvSciSync = 6, cudaExternalSemaphoreHandleTypeKeyedMutex = 7, cudaExternalSemaphoreHandleTypeKeyedMutexKmt = 8, cudaExternalSemaphoreHandleTypeTimelineSemaphoreFd = 9, cudaExternalSemaphoreHandleTypeTimelineSemaphoreWin32 = 10 }; struct __device_builtin__ cudaExternalSemaphoreHandleDesc { enum cudaExternalSemaphoreHandleType type; union { int fd; struct { void *handle; const void *name; } win32; const void* nvSciSyncObj; } handle; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams_v1 { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; } params; unsigned int flags; }; struct __device_builtin__ cudaExternalSemaphoreSignalParams{ struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; } keyedMutex; unsigned int reserved[12]; } params; unsigned int flags; unsigned int reserved[16]; }; struct __device_builtin__ cudaExternalSemaphoreWaitParams { struct { struct { unsigned long long value; } fence; union { void *fence; unsigned long long reserved; } nvSciSync; struct { unsigned long long key; unsigned int timeoutMs; } keyedMutex; unsigned int reserved[10]; } params; unsigned int flags; unsigned int reserved[16]; }; typedef __device_builtin__ enum cudaError cudaError_t; typedef __device_builtin__ struct CUstream_st *cudaStream_t; typedef __device_builtin__ struct CUevent_st *cudaEvent_t; typedef __device_builtin__ struct cudaGraphicsResource *cudaGraphicsResource_t; typedef __device_builtin__ struct CUexternalMemory_st *cudaExternalMemory_t; typedef __device_builtin__ struct CUexternalSemaphore_st *cudaExternalSemaphore_t; typedef __device_builtin__ struct CUgraph_st *cudaGraph_t; typedef __device_builtin__ struct CUgraphNode_st *cudaGraphNode_t; typedef __device_builtin__ struct CUuserObject_st *cudaUserObject_t; typedef __device_builtin__ struct CUfunc_st *cudaFunction_t; typedef __device_builtin__ struct CUkern_st *cudaKernel_t; typedef __device_builtin__ struct CUmemPoolHandle_st *cudaMemPool_t; enum __device_builtin__ cudaCGScope { cudaCGScopeInvalid = 0, cudaCGScopeGrid = 1, cudaCGScopeMultiGrid = 2 }; struct __device_builtin__ cudaLaunchParams { void *func; dim3 gridDim; dim3 blockDim; void **args; size_t sharedMem; cudaStream_t stream; }; struct __device_builtin__ cudaKernelNodeParams { void* func; dim3 gridDim; dim3 blockDim; unsigned int sharedMemBytes; void **kernelParams; void **extra; }; struct __device_builtin__ cudaExternalSemaphoreSignalNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreSignalParams* paramsArray; unsigned int numExtSems; }; struct __device_builtin__ cudaExternalSemaphoreWaitNodeParams { cudaExternalSemaphore_t* extSemArray; const struct cudaExternalSemaphoreWaitParams* paramsArray; unsigned int numExtSems; }; enum __device_builtin__ cudaGraphNodeType { cudaGraphNodeTypeKernel = 0x00, cudaGraphNodeTypeMemcpy = 0x01, cudaGraphNodeTypeMemset = 0x02, cudaGraphNodeTypeHost = 0x03, cudaGraphNodeTypeGraph = 0x04, cudaGraphNodeTypeEmpty = 0x05, cudaGraphNodeTypeWaitEvent = 0x06, cudaGraphNodeTypeEventRecord = 0x07, cudaGraphNodeTypeExtSemaphoreSignal = 0x08, cudaGraphNodeTypeExtSemaphoreWait = 0x09, cudaGraphNodeTypeMemAlloc = 0x0a, cudaGraphNodeTypeMemFree = 0x0b, cudaGraphNodeTypeCount }; typedef struct CUgraphExec_st* cudaGraphExec_t; enum __device_builtin__ cudaGraphExecUpdateResult { cudaGraphExecUpdateSuccess = 0x0, cudaGraphExecUpdateError = 0x1, cudaGraphExecUpdateErrorTopologyChanged = 0x2, cudaGraphExecUpdateErrorNodeTypeChanged = 0x3, cudaGraphExecUpdateErrorFunctionChanged = 0x4, cudaGraphExecUpdateErrorParametersChanged = 0x5, cudaGraphExecUpdateErrorNotSupported = 0x6, cudaGraphExecUpdateErrorUnsupportedFunctionChange = 0x7, cudaGraphExecUpdateErrorAttributesChanged = 0x8 }; typedef __device_builtin__ enum cudaGraphInstantiateResult { cudaGraphInstantiateSuccess = 0, cudaGraphInstantiateError = 1, cudaGraphInstantiateInvalidStructure = 2, cudaGraphInstantiateNodeOperationNotSupported = 3, cudaGraphInstantiateMultipleDevicesNotSupported = 4 } cudaGraphInstantiateResult; typedef __device_builtin__ struct cudaGraphInstantiateParams_st { unsigned long long flags; cudaStream_t uploadStream; cudaGraphNode_t errNode_out; cudaGraphInstantiateResult result_out; } cudaGraphInstantiateParams; typedef __device_builtin__ struct cudaGraphExecUpdateResultInfo_st { enum cudaGraphExecUpdateResult result; cudaGraphNode_t errorNode; cudaGraphNode_t errorFromNode; } cudaGraphExecUpdateResultInfo; enum __device_builtin__ cudaGetDriverEntryPointFlags { cudaEnableDefault = 0x0, cudaEnableLegacyStream = 0x1, cudaEnablePerThreadDefaultStream = 0x2 }; enum __device_builtin__ cudaDriverEntryPointQueryResult { cudaDriverEntryPointSuccess = 0, cudaDriverEntryPointSymbolNotFound = 1, cudaDriverEntryPointVersionNotSufficent = 2 }; enum __device_builtin__ cudaGraphDebugDotFlags { cudaGraphDebugDotFlagsVerbose = 1<<0, cudaGraphDebugDotFlagsKernelNodeParams = 1<<2, cudaGraphDebugDotFlagsMemcpyNodeParams = 1<<3, cudaGraphDebugDotFlagsMemsetNodeParams = 1<<4, cudaGraphDebugDotFlagsHostNodeParams = 1<<5, cudaGraphDebugDotFlagsEventNodeParams = 1<<6, cudaGraphDebugDotFlagsExtSemasSignalNodeParams = 1<<7, cudaGraphDebugDotFlagsExtSemasWaitNodeParams = 1<<8, cudaGraphDebugDotFlagsKernelNodeAttributes = 1<<9, cudaGraphDebugDotFlagsHandles = 1<<10 }; enum __device_builtin__ cudaGraphInstantiateFlags { cudaGraphInstantiateFlagAutoFreeOnLaunch = 1 , cudaGraphInstantiateFlagUpload = 2 , cudaGraphInstantiateFlagDeviceLaunch = 4 , cudaGraphInstantiateFlagUseNodePriority = 8 }; typedef __device_builtin__ enum cudaLaunchMemSyncDomain { cudaLaunchMemSyncDomainDefault = 0, cudaLaunchMemSyncDomainRemote = 1 } cudaLaunchMemSyncDomain; typedef __device_builtin__ struct cudaLaunchMemSyncDomainMap_st { unsigned char default_; unsigned char remote; } cudaLaunchMemSyncDomainMap; typedef __device_builtin__ enum cudaLaunchAttributeID { cudaLaunchAttributeIgnore = 0 , cudaLaunchAttributeAccessPolicyWindow = 1 , cudaLaunchAttributeCooperative = 2 , cudaLaunchAttributeSynchronizationPolicy = 3 , cudaLaunchAttributeClusterDimension = 4 , cudaLaunchAttributeClusterSchedulingPolicyPreference = 5 , cudaLaunchAttributeProgrammaticStreamSerialization = 6 , cudaLaunchAttributeProgrammaticEvent = 7 , cudaLaunchAttributePriority = 8 , cudaLaunchAttributeMemSyncDomainMap = 9 , cudaLaunchAttributeMemSyncDomain = 10 } cudaLaunchAttributeID; typedef __device_builtin__ union cudaLaunchAttributeValue { char pad[64]; struct cudaAccessPolicyWindow accessPolicyWindow; int cooperative; enum cudaSynchronizationPolicy syncPolicy; struct { unsigned int x; unsigned int y; unsigned int z; } clusterDim; enum cudaClusterSchedulingPolicy clusterSchedulingPolicyPreference; int programmaticStreamSerializationAllowed; struct { cudaEvent_t event; int flags; int triggerAtBlockStart; } programmaticEvent; int priority; cudaLaunchMemSyncDomainMap memSyncDomainMap; cudaLaunchMemSyncDomain memSyncDomain; } cudaLaunchAttributeValue; typedef __device_builtin__ struct cudaLaunchAttribute_st { cudaLaunchAttributeID id; char pad[8 - sizeof(cudaLaunchAttributeID)]; cudaLaunchAttributeValue val; } cudaLaunchAttribute; typedef __device_builtin__ struct cudaLaunchConfig_st { dim3 gridDim; dim3 blockDim; size_t dynamicSmemBytes; cudaStream_t stream; cudaLaunchAttribute *attrs; unsigned int numAttrs; } cudaLaunchConfig_t; #define cudaStreamAttrID cudaLaunchAttributeID #define cudaStreamAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaStreamAttributeSynchronizationPolicy cudaLaunchAttributeSynchronizationPolicy #define cudaStreamAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaStreamAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaStreamAttributePriority cudaLaunchAttributePriority #define cudaStreamAttrValue cudaLaunchAttributeValue #define cudaKernelNodeAttrID cudaLaunchAttributeID #define cudaKernelNodeAttributeAccessPolicyWindow cudaLaunchAttributeAccessPolicyWindow #define cudaKernelNodeAttributeCooperative cudaLaunchAttributeCooperative #define cudaKernelNodeAttributePriority cudaLaunchAttributePriority #define cudaKernelNodeAttributeClusterDimension cudaLaunchAttributeClusterDimension #define cudaKernelNodeAttributeClusterSchedulingPolicyPreference cudaLaunchAttributeClusterSchedulingPolicyPreference #define cudaKernelNodeAttributeMemSyncDomainMap cudaLaunchAttributeMemSyncDomainMap #define cudaKernelNodeAttributeMemSyncDomain cudaLaunchAttributeMemSyncDomain #define cudaKernelNodeAttrValue cudaLaunchAttributeValue #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_DRIVER_TYPES_H__ #endif #undef __CUDA_DEPRECATED #endif #define cudaTextureType1D 0x01 #define cudaTextureType2D 0x02 #define cudaTextureType3D 0x03 #define cudaTextureTypeCubemap 0x0C #define cudaTextureType1DLayered 0xF1 #define cudaTextureType2DLayered 0xF2 #define cudaTextureTypeCubemapLayered 0xFC enum __device_builtin__ cudaTextureAddressMode { cudaAddressModeWrap = 0, cudaAddressModeClamp = 1, cudaAddressModeMirror = 2, cudaAddressModeBorder = 3 }; enum __device_builtin__ cudaTextureFilterMode { cudaFilterModePoint = 0, cudaFilterModeLinear = 1 }; enum __device_builtin__ cudaTextureReadMode { cudaReadModeElementType = 0, cudaReadModeNormalizedFloat = 1 }; struct __device_builtin__ cudaTextureDesc { enum cudaTextureAddressMode addressMode[3]; enum cudaTextureFilterMode filterMode; enum cudaTextureReadMode readMode; int sRGB; float borderColor[4]; int normalizedCoords; unsigned int maxAnisotropy; enum cudaTextureFilterMode mipmapFilterMode; float mipmapLevelBias; float minMipmapLevelClamp; float maxMipmapLevelClamp; int disableTrilinearOptimization; int seamlessCubemap; }; typedef __device_builtin__ unsigned long long cudaTextureObject_t; #endif #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif extern "C" { #define __DEVICE_FUNCTIONS_DECL__ __host__ __device__ #if !defined(_MSC_VER) #define __CUDA_MATH_CRTIMP #else #if _MSC_VER < 1900 #define __CUDA_MATH_CRTIMP _CRTIMP #else #define __CUDA_MATH_CRTIMP _ACRTIMP #endif #endif #if defined(__ANDROID__) && (__ANDROID_API__ <= 20) && !defined(__aarch64__) static __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ int abs(int); static __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ long int labs(long int); static __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ long long int llabs(long long int); #else #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ int __cdecl abs(int a) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ long int __cdecl labs(long int a) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __cudart_builtin__ long long int llabs(long long int a) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl fabs(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fabsf(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int min(const int a, const int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int umin(const unsigned int a, const unsigned int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llmin(const long long int a, const long long int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int ullmin(const unsigned long long int a, const unsigned long long int b); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fminf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl fminf(float x, float y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double fmin(double x, double y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl fmin(double x, double y); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int max(const int a, const int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned int umax(const unsigned int a, const unsigned int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llmax(const long long int a, const long long int b); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ unsigned long long int ullmax(const unsigned long long int a, const unsigned long long int b); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fmaxf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl fmaxf(float x, float y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double fmax(double, double) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl fmax(double, double); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl sin(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl cos(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void sincos(double x, double *sptr, double *cptr) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void sincosf(float x, float *sptr, float *cptr) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl tan(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl sqrt(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double rsqrt(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float rsqrtf(float x); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double log2(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl log2(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double exp2(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl exp2(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float exp2f(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl exp2f(float x); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double exp10(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float exp10f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double expm1(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl expm1(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float expm1f(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl expm1f(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float log2f(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl log2f(float x); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl log10(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl log(double x) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double log1p(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl log1p(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float log1pf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl log1pf(float x); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl floor(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl exp(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl cosh(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl sinh(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl tanh(double x) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double acosh(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl acosh(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float acoshf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl acoshf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double asinh(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl asinh(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float asinhf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl asinhf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double atanh(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl atanh(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float atanhf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl atanhf(float x); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl ldexp(double x, int exp) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float ldexpf(float x, int exp) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double logb(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl logb(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float logbf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl logbf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int ilogb(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP int __cdecl ilogb(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int ilogbf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP int __cdecl ilogbf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double scalbn(double x, int n) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl scalbn(double x, int n); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float scalbnf(float x, int n) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl scalbnf(float x, int n); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double scalbln(double x, long int n) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl scalbln(double x, long int n); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float scalblnf(float x, long int n) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl scalblnf(float x, long int n); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl frexp(double x, int *nptr) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float frexpf(float x, int *nptr) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double round(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl round(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float roundf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl roundf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long int lround(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lround(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long int lroundf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lroundf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llround(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llround(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llroundf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llroundf(float x); #endif #if defined(__CUDA_ARCH__) || defined(__DOXYGEN_ONLY__) #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double rint(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl rint(double x); #endif #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float rintf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl rintf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long int lrint(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lrint(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long int lrintf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lrintf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llrint(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llrint(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ long long int llrintf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llrintf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double nearbyint(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl nearbyint(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float nearbyintf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl nearbyintf(float x); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl ceil(double x) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double trunc(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl trunc(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float truncf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl truncf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double fdim(double x, double y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl fdim(double x, double y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fdimf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl fdimf(float x, float y); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl atan2(double y, double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl atan(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl acos(double x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl asin(double x) __THROW; #if defined(_WIN32) #if defined(_MSC_VER) && _MSC_VER < 1900 static __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __CRTDECL hypot(double x, double y); #else extern _ACRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl hypot(double x, double y); #endif #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double hypot(double x, double y) __THROW; #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __device__ __device_builtin__ double rhypot(double x, double y) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if defined(_WIN32) static __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __CRTDECL hypotf(float x, float y); #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float hypotf(float x, float y) __THROW; #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __device__ __device_builtin__ float rhypotf(float x, float y) __THROW; extern __device__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl norm3d(double a, double b, double c) __THROW; extern __device__ __device_builtin__ double rnorm3d(double a, double b, double c) __THROW; extern __device__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl norm4d(double a, double b, double c, double d) __THROW; extern __device__ __device_builtin__ double rnorm4d(double a, double b, double c, double d) __THROW; extern "C++" __device__ __device_builtin__ double norm(int dim, double const * p) __THROW; extern __device__ __device_builtin__ double rnorm(int dim, double const * p) __THROW; extern __device__ __device_builtin__ float rnormf(int dim, float const * p) __THROW; extern "C++" __device__ __device_builtin__ float normf(int dim, float const * p) __THROW; extern __device__ __device_builtin__ float norm3df(float a, float b, float c) __THROW; extern __device__ __device_builtin__ float rnorm3df(float a, float b, float c) __THROW; extern __device__ __device_builtin__ float norm4df(float a, float b, float c, float d) __THROW; extern __device__ __device_builtin__ float rnorm4df(float a, float b, float c, float d) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double cbrt(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl cbrt(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float cbrtf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl cbrtf(float x); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double rcbrt(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float rcbrtf(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double sinpi(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float sinpif(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double cospi(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float cospif(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void sincospi(double x, double *sptr, double *cptr); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ void sincospif(float x, float *sptr, float *cptr); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl pow(double x, double y) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl modf(double x, double *iptr) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double __cdecl fmod(double x, double y) __THROW; #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double remainder(double x, double y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl remainder(double x, double y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float remainderf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl remainderf(float x, float y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double remquo(double x, double y, int *quo) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl remquo(double x, double y, int *quo); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float remquof(float x, float y, int *quo) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl remquof(float x, float y, int *quo); #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl j0(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float j0f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl j1(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float j1f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl jn(int n, double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float jnf(int n, float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl y0(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float y0f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl y1(double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float y1f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl yn(int n, double x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float ynf(int n, float x) __THROW; extern __device__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl cyl_bessel_i0(double x) __THROW; extern __device__ __device_builtin__ float cyl_bessel_i0f(float x) __THROW; extern __device__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl cyl_bessel_i1(double x) __THROW; extern __device__ __device_builtin__ float cyl_bessel_i1f(float x) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double erf(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl erf(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float erff(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl erff(float x); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double erfinv(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float erfinvf(float x); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double erfc(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl erfc(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float erfcf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl erfcf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double lgamma(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl lgamma(double x); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double erfcinv(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float erfcinvf(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double normcdfinv(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float normcdfinvf(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double normcdf(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float normcdff(float x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double erfcx(double x); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float erfcxf(float x); #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float lgammaf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl lgammaf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double tgamma(double x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl tgamma(double x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float tgammaf(float x) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl tgammaf(float x); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double copysign(double x, double y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl copysign(double x, double y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float copysignf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl copysignf(float x, float y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double nextafter(double x, double y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl nextafter(double x, double y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float nextafterf(float x, float y) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl nextafterf(float x, float y); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double nan(const char *tagp) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl nan(const char *tagp); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float nanf(const char *tagp) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl nanf(const char *tagp); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isinff(float) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isnanf(float) __THROW; #if defined(__APPLE__) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isfinited(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isfinitef(float) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __signbitd(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isnand(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isinfd(double) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __finite(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __finitef(float) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __signbit(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isnan(double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isinf(double) __THROW; #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __signbitf(float) __THROW; #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double fma(double x, double y, double z) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl fma(double x, double y, double z); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fmaf(float x, float y, float z) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl fmaf(float x, float y, float z); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __signbitl(long double) __THROW; #if defined(__APPLE__) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isfinite(long double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isinf(long double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isnan(long double) __THROW; #else extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __finitel(long double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isinfl(long double) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int __isnanl(long double) __THROW; #endif #if defined(_WIN32) && defined(_M_AMD64) extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl acosf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl asinf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl atanf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl atan2f(float, float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl cosf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl sinf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl tanf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl coshf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl sinhf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl tanhf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl expf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl logf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl log10f(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl modff(float, float*) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl powf(float, float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl sqrtf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl ceilf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl floorf(float) __THROW; extern __CUDA_MATH_CRTIMP __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float __cdecl fmodf(float, float) __THROW; #else #if defined(__QNX__) && !defined(_LIBCPP_VERSION) namespace std { #endif extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float acosf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float asinf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float atanf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float atan2f(float y, float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float cosf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float sinf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float tanf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float coshf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float sinhf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float tanhf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float logf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float expf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float log10f(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float modff(float x, float *iptr) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float powf(float x, float y) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float sqrtf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float ceilf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float floorf(float x) __THROW; extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float fmodf(float x, float y) __THROW; #if defined(__QNX__) extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float _FLog(float, int); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float _FCosh(float, float); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float _FSinh(float, float); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float _FSinx(float, unsigned int, int); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int _FDsign(float); extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ int _Dsign(double); #endif #if defined(__QNX__) && !defined(_LIBCPP_VERSION) } #endif #endif } #if !defined(__CUDACC_RTC__) #include <math.h> #include <stdlib.h> #ifndef __CUDA_INTERNAL_SKIP_CPP_HEADERS__ #include <cmath> #include <cstdlib> #endif #endif #if defined(__CUDACC_RTC__) __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(long double x); #elif defined(__GNUC__) #undef signbit #undef isfinite #undef isnan #undef isinf #if defined(__APPLE__) __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(float x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(double x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(long double x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(float x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(double x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(long double x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(double x) throw(); #if !defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 7000 __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(float x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(long double x); #else template <typename T> __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool __libcpp_isnan(T) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isnan(float x) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isnan(long double x) _NOEXCEPT; #endif __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(double x) throw(); #if !defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 7000 __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(float x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(long double x); #else template <typename T> __cudart_builtin__ __DEVICE_FUNCTIONS_DECL__ bool __libcpp_isinf(T) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isinf(float x) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isinf(long double x) _NOEXCEPT; #endif #else #if ((defined _GLIBCXX_MATH_H) && _GLIBCXX_MATH_H) && (__cplusplus >= 201103L) #if !defined(_NVHPC_CUDA) namespace std { __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool signbit(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool signbit(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool signbit(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isfinite(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isfinite(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isfinite(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isnan(float x); #if _GLIBCXX_HAVE_OBSOLETE_ISNAN && !_GLIBCXX_NO_OBSOLETE_ISINF_ISNAN_DYNAMIC __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(double x) throw(); #else __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isnan(double x); #endif __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isnan(long double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isinf(float x); #if _GLIBCXX_HAVE_OBSOLETE_ISINF && !_GLIBCXX_NO_OBSOLETE_ISINF_ISNAN_DYNAMIC __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(double x) throw(); #else __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isinf(double x); #endif __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ constexpr bool isinf(long double x); } #endif #else #if defined(__QNX__) #if (__QNX__) && !defined(_LIBCPP_VERSION) namespace std { __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool signbit(float x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool signbit(double x); __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool signbit(long double x); } #else static __inline__ __DEVICE_FUNCTIONS_DECL__ bool signbit(const float x); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool signbit(const double x); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool signbit(const long double x); #endif static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isfinite(const float a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isfinite(const double a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isfinite(const long double a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isnan(const float a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isnan(const double a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isnan(const long double a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isinf(const float a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isinf(const double a); static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isinf(const long double a); #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(const float x); #if defined(__ICC) __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(const double x) throw(); #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(const double x); #endif __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int signbit(const long double x); __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(const float x); #if defined(__ICC) __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(const double x) throw(); #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(const double x); #endif __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isfinite(const long double x); #if (defined(__ANDROID__) || defined(__HORIZON__)) && _LIBCPP_VERSION >= 8000 template <typename T> __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool __libcpp_isnan(T) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isnan(float x) _NOEXCEPT; #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(float x); #endif #if defined(__ANDROID__) || defined(__HORIZON__) #if !defined(_LIBCPP_VERSION) __forceinline__ #endif #if _LIBCPP_VERSION >= 7000 #ifdef _LIBCPP_PREFERRED_OVERLOAD _LIBCPP_INLINE_VISIBILITY _LIBCPP_PREFERRED_OVERLOAD __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isnan(double x) _NOEXCEPT; #endif #else __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(double x); #endif #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(double x) throw(); #endif #if (defined(__ANDROID__) || defined(__HORIZON__)) && _LIBCPP_VERSION >= 8000 inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isnan(long double x) _NOEXCEPT; #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isnan(long double x); #endif #if (defined(__ANDROID__) || defined(__HORIZON__)) && _LIBCPP_VERSION >= 8000 static __inline__ __cudart_builtin__ __DEVICE_FUNCTIONS_DECL__ unsigned __FLOAT_BITS(float __f); static __inline__ __cudart_builtin__ __DEVICE_FUNCTIONS_DECL__ unsigned long long __DOUBLE_BITS(double __f); template <typename T> __cudart_builtin__ __DEVICE_FUNCTIONS_DECL__ bool __libcpp_isinf(T) _NOEXCEPT; inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isinf(float x) _NOEXCEPT; #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(float x); #endif #if defined(__ANDROID__) || defined(__HORIZON__) #if !defined(_LIBCPP_VERSION) __forceinline__ #endif #if _LIBCPP_VERSION >= 7000 #ifdef _LIBCPP_PREFERRED_OVERLOAD _LIBCPP_INLINE_VISIBILITY _LIBCPP_PREFERRED_OVERLOAD __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isinf(double x) _NOEXCEPT; #endif #else __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(double x); #endif #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(double x) throw(); #endif #if (defined(__ANDROID__) || defined(__HORIZON__)) && _LIBCPP_VERSION >= 8000 inline _LIBCPP_INLINE_VISIBILITY __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool isinf(long double x) _NOEXCEPT; #else __forceinline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ int isinf(long double x); #endif #endif #endif #endif #if !defined(_LIBCPP_VERSION) #if defined(__clang__) #if __has_include(<ext/random>) #define __NV_GLIBCXX_VERSION 40800 #endif #endif #if !defined(__NV_GLIBCXX_VERSION) #define __NV_GLIBCXX_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) #endif #endif #if !defined(__HORIZON__) || !defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 3800 #if defined(__arm__) && !defined(_STLPORT_VERSION) && !_GLIBCXX_USE_C99 #if !defined(__ANDROID__) || (defined(__NV_GLIBCXX_VERSION) && __NV_GLIBCXX_VERSION < 40800) #if defined(__QNX__) namespace std { __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs (long long int a); } #elif defined(__HORIZON__) #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER) #pragma GCC system_header #endif _LIBCPP_BEGIN_NAMESPACE_STD __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs (long long int a) throw(); _LIBCPP_END_NAMESPACE_STD #else static __inline__ __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs(long long int a); #endif #endif #endif #endif #if defined(__NV_GLIBCXX_VERSION) && __NV_GLIBCXX_VERSION < 40800 && !defined(__ibmxl__) #if !defined(_STLPORT_VERSION) namespace __gnu_cxx { #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs(long long int a); #if !defined(_STLPORT_VERSION) } #endif #endif namespace std { template<typename T> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ T __pow_helper(T, int); template<typename T> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ T __cmath_power(T, unsigned int); } using std::abs; using std::fabs; using std::ceil; using std::floor; using std::sqrt; #if !defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 3800 using std::pow; #endif using std::log; using std::log10; using std::fmod; using std::modf; using std::exp; using std::frexp; using std::ldexp; using std::asin; using std::sin; using std::sinh; using std::acos; using std::cos; using std::cosh; using std::atan; using std::atan2; using std::tan; using std::tanh; #elif defined(_WIN32) extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __CUDA_MATH_CRTIMP double __cdecl _hypot(double x, double y); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __CUDA_MATH_CRTIMP float __cdecl _hypotf(float x, float y); #if (!defined(_MSC_VER) || _MSC_VER < 1800) static __inline__ __DEVICE_FUNCTIONS_DECL__ int signbit(long double a); #else #if _MSC_VER >= 1900 #define __SIGNBIT_THROW throw() #else #define __SIGNBIT_THROW #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ bool signbit(long double) __SIGNBIT_THROW; extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __device_builtin__ __CUDA_MATH_CRTIMP int _ldsign(long double); #undef __SIGNBIT_THROW #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE signbit(double a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool #if _MSC_VER >= 1900 #define __SIGNBIT_THROW throw() #else #define __SIGNBIT_THROW #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __RETURN_TYPE signbit(double) __SIGNBIT_THROW; extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __device_builtin__ __CUDA_MATH_CRTIMP int _dsign(double); #undef __RETURN_TYPE #undef __SIGNBIT_THROW #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE signbit(float a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool #if _MSC_VER >= 1900 #define __SIGNBIT_THROW throw() #else #define __SIGNBIT_THROW #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __RETURN_TYPE signbit(float) __SIGNBIT_THROW; extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ __device_builtin__ __CUDA_MATH_CRTIMP int _fdsign(float); #undef __RETURN_TYPE #undef __SIGNBIT_THROW #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) static __inline__ __DEVICE_FUNCTIONS_DECL__ int isinf(long double a); #else static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isinf(long double a); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isinf(double a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isinf(double a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isinf(float a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isinf(float a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) static __inline__ __DEVICE_FUNCTIONS_DECL__ int isnan(long double a); #else static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isnan(long double a); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isnan(double a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isnan(double a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isnan(float a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isnan(float a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) static __inline__ __DEVICE_FUNCTIONS_DECL__ int isfinite(long double a); #else static __inline__ __DEVICE_FUNCTIONS_DECL__ bool isfinite(long double a); #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isfinite(double a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isfinite(double a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #define __RETURN_TYPE int static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isfinite(float a); #undef __RETURN_TYPE #else #define __RETURN_TYPE bool static __inline__ __DEVICE_FUNCTIONS_DECL__ __RETURN_TYPE isfinite(float a); #undef __RETURN_TYPE #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) template<class T> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ T _Pow_int(T, int); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs(long long int); #else template<class T> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ T _Pow_int(T, int) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs(long long int) throw(); #endif #endif #if __cplusplus >= 201103L #define __NV_NOEXCEPT noexcept #else #define __NV_NOEXCEPT throw() #endif #if defined(_LIBCPP_VERSION) && defined(_LIBCPP_BEGIN_NAMESPACE_STD) && !defined(_STLPORT_VERSION) #if defined(__clang__) #pragma clang diagnostic push #pragma clang diagnostic ignored "-Wc++11-extensions" #endif #if _LIBCPP_VERSION < 3800 _LIBCPP_BEGIN_NAMESPACE_STD #endif #elif defined(__GNUC__) && !defined(_STLPORT_VERSION) namespace std { #endif #if defined(__CUDACC_RTC__) || defined(__GNUC__) #if defined(__CUDACC_RTC__) || \ (defined(__NV_GLIBCXX_VERSION) && __NV_GLIBCXX_VERSION >= 40800) || \ defined(__ibmxl__) extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int abs(long long int); #endif #endif #if defined(__CUDACC_RTC__) || \ (!defined(_MSC_VER) || _MSC_VER < 1800) && \ (!defined(_LIBCPP_VERSION) || (_LIBCPP_VERSION < 1101)) extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long int __cdecl abs(long int); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl abs(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ double __cdecl abs(double); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl fabs(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl ceil(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl floor(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sqrt(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl pow(float, float); #if !defined(__QNX__) #if defined(__GNUC__) && __cplusplus >= 201103L && !defined(_LIBCPP_VERSION) template<typename _Tp, typename _Up> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ typename __gnu_cxx::__promote_2<_Tp, _Up>::__type pow(_Tp, _Up); #else extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl pow(float, int); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ double __cdecl pow(double, int); #endif #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl log(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl log10(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl fmod(float, float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl modf(float, float*); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl exp(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl frexp(float, int*); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl ldexp(float, int); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl asin(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sin(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sinh(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl acos(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl cos(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl cosh(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl atan(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl atan2(float, float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl tan(float); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl tanh(float); #else extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long int __cdecl abs(long int) throw(); #if defined(_LIBCPP_VERSION) extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ long long int __cdecl abs(long long int) throw(); #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl abs(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ double __cdecl abs(double) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl fabs(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl ceil(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl floor(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sqrt(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl pow(float, float) throw(); #if defined(_LIBCPP_VERSION) #if (defined (__ANDROID__) || defined(__HORIZON__)) && (_LIBCPP_VERSION >= 9000) template <class _A1, class _A2> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ #if _LIBCPP_VERSION >= 14000 typename std::__enable_if_t #else typename std::_EnableIf #endif < std::is_arithmetic<_A1>::value && std::is_arithmetic<_A2>::value, std::__promote<_A1, _A2> >::type pow(_A1 __lcpp_x, _A2 __lcpp_y) __NV_NOEXCEPT; #elif (defined(__APPLE__) && __clang_major__ >= 7) || _LIBCPP_VERSION >= 3800 || defined(__QNX__) template <class _Tp, class _Up> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ typename std::__lazy_enable_if < std::is_arithmetic<_Tp>::value && std::is_arithmetic<_Up>::value, std::__promote<_Tp, _Up> >::type pow(_Tp __x, _Up __y) __NV_NOEXCEPT; #else template <class _Tp, class _Up> extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ typename enable_if < std::is_arithmetic<_Tp>::value && std::is_arithmetic<_Up>::value, typename std::__promote<_Tp, _Up>::type >::type pow(_Tp __x, _Up __y) __NV_NOEXCEPT; #endif #else #if !(defined(__GNUC__) && __cplusplus >= 201103L) #if (defined(_MSC_VER) && (_MSC_VER >= 1928)) && !(defined __CUDA_INTERNAL_SKIP_CPP_HEADERS__) template <class _Ty1, class _Ty2, ::std:: enable_if_t< ::std:: is_arithmetic_v<_Ty1> && ::std:: is_arithmetic_v<_Ty2>, int> > [[nodiscard]] __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ ::std:: _Common_float_type_t<_Ty1, _Ty2> __cdecl pow(_Ty1 _Left, _Ty2 _Right) noexcept; #else extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl pow(float, int) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ double __cdecl pow(double, int) throw(); #endif #endif #endif extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl log(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl log10(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl fmod(float, float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl modf(float, float*) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl exp(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl frexp(float, int*) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl ldexp(float, int) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl asin(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sin(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl sinh(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl acos(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl cos(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl cosh(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl atan(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl atan2(float, float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl tan(float) throw(); extern __DEVICE_FUNCTIONS_DECL__ __cudart_builtin__ float __cdecl tanh(float) throw(); #endif #if defined(_LIBCPP_VERSION) && defined(_LIBCPP_END_NAMESPACE_STD) && !defined(_STLPORT_VERSION) #if _LIBCPP_VERSION < 3800 _LIBCPP_END_NAMESPACE_STD #endif #if defined(__clang__) #pragma clang diagnostic pop #endif #elif defined(__GNUC__) && !defined(_STLPORT_VERSION) } #endif #undef __DEVICE_FUNCTIONS_DECL__ #undef __NV_NOEXCEPT #if defined(__CUDACC_RTC__) #define __MATH_FUNCTIONS_DECL__ __host__ __device__ #define __MATH_FUNCTIONS_DEVICE_DECL__ __device__ #else #define __MATH_FUNCTIONS_DECL__ static inline __host__ __device__ __cudart_builtin__ #define __MATH_FUNCTIONS_DEVICE_DECL__ static inline __device__ __cudart_builtin__ #endif #if (!defined(_MSC_VER) || _MSC_VER < 1800) #if defined(__QNX__) || (defined(_LIBCPP_VERSION) && _LIBCPP_VERSION >= 3800) #if defined(__QNX__) && (!defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 8000) #if defined(_LIBCPP_VERSION) #define __NV_NOEXCEPT _NOEXCEPT _LIBCPP_BEGIN_NAMESPACE_STD #else #define __NV_NOEXCEPT namespace std { __host__ __device__ __cudart_builtin__ int ilogbf(float a); #endif #else #define __NV_NOEXCEPT _NOEXCEPT #endif __host__ __device__ __cudart_builtin__ float logb(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ int ilogb(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float scalbn(float a, int b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float scalbln(float a, long int b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float exp2(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float expm1(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float log2(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float log1p(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float acosh(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float asinh(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float atanh(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float hypot(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float cbrt(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float erf(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float erfc(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float lgamma(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float tgamma(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float copysign(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float nextafter(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float remainder(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float remquo(float a, float b, int *quo) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float round(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ long int lround(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ long long int llround(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float trunc(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float rint(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ long int lrint(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ long long int llrint(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float nearbyint(float a) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float fdim(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float fma(float a, float b, float c) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float fmax(float a, float b) __NV_NOEXCEPT; __host__ __device__ __cudart_builtin__ float fmin(float a, float b) __NV_NOEXCEPT; #if defined(__QNX__) && (!defined(_LIBCPP_VERSION) || _LIBCPP_VERSION < 8000) #if defined(_LIBCPP_VERSION) _LIBCPP_END_NAMESPACE_STD using _VSTD::logb; using _VSTD::ilogb; using _VSTD::scalbn; using _VSTD::scalbln; using _VSTD::exp2; using _VSTD::expm1; using _VSTD::log2; using _VSTD::log1p; using _VSTD::acosh; using _VSTD::asinh; using _VSTD::atanh; using _VSTD::hypot; using _VSTD::cbrt; using _VSTD::erf; using _VSTD::erfc; using _VSTD::lgamma; using _VSTD::tgamma; using _VSTD::copysign; using _VSTD::nextafter; using _VSTD::remainder; using _VSTD::remquo; using _VSTD::round; using _VSTD::lround; using _VSTD::llround; using _VSTD::trunc; using _VSTD::rint; using _VSTD::lrint; using _VSTD::llrint; using _VSTD::nearbyint; using _VSTD::fdim; using _VSTD::fma; using _VSTD::fmax; using _VSTD::fmin; #else } #endif #endif #undef __NV_NOEXCEPT #else #if ((defined _GLIBCXX_MATH_H) && _GLIBCXX_MATH_H) && (__cplusplus >= 201103L) namespace std { __host__ __device__ __cudart_builtin__ constexpr float logb(float a); __host__ __device__ __cudart_builtin__ constexpr int ilogb(float a); __host__ __device__ __cudart_builtin__ constexpr float scalbn(float a, int b); __host__ __device__ __cudart_builtin__ constexpr float scalbln(float a, long int b); __host__ __device__ __cudart_builtin__ constexpr float exp2(float a); __host__ __device__ __cudart_builtin__ constexpr float expm1(float a); __host__ __device__ __cudart_builtin__ constexpr float log2(float a); __host__ __device__ __cudart_builtin__ constexpr float log1p(float a); __host__ __device__ __cudart_builtin__ constexpr float acosh(float a); __host__ __device__ __cudart_builtin__ constexpr float asinh(float a); __host__ __device__ __cudart_builtin__ constexpr float atanh(float a); __host__ __device__ __cudart_builtin__ constexpr float hypot(float a, float b); __host__ __device__ __cudart_builtin__ constexpr float cbrt(float a); __host__ __device__ __cudart_builtin__ constexpr float erf(float a); __host__ __device__ __cudart_builtin__ constexpr float erfc(float a); __host__ __device__ __cudart_builtin__ constexpr float lgamma(float a); __host__ __device__ __cudart_builtin__ constexpr float tgamma(float a); __host__ __device__ __cudart_builtin__ constexpr float copysign(float a, float b); __host__ __device__ __cudart_builtin__ constexpr float nextafter(float a, float b); __host__ __device__ __cudart_builtin__ constexpr float remainder(float a, float b); __host__ __device__ __cudart_builtin__ float remquo(float a, float b, int *quo); __host__ __device__ __cudart_builtin__ constexpr float round(float a); __host__ __device__ __cudart_builtin__ constexpr long int lround(float a); __host__ __device__ __cudart_builtin__ constexpr long long int llround(float a); __host__ __device__ __cudart_builtin__ constexpr float trunc(float a); __host__ __device__ __cudart_builtin__ constexpr float rint(float a); __host__ __device__ __cudart_builtin__ constexpr long int lrint(float a); __host__ __device__ __cudart_builtin__ constexpr long long int llrint(float a); __host__ __device__ __cudart_builtin__ constexpr float nearbyint(float a); __host__ __device__ __cudart_builtin__ constexpr float fdim(float a, float b); __host__ __device__ __cudart_builtin__ constexpr float fma(float a, float b, float c); __host__ __device__ __cudart_builtin__ constexpr float fmax(float a, float b); __host__ __device__ __cudart_builtin__ constexpr float fmin(float a, float b); } #else __MATH_FUNCTIONS_DECL__ float logb(float a); __MATH_FUNCTIONS_DECL__ int ilogb(float a); __MATH_FUNCTIONS_DECL__ float scalbn(float a, int b); __MATH_FUNCTIONS_DECL__ float scalbln(float a, long int b); __MATH_FUNCTIONS_DECL__ float exp2(float a); __MATH_FUNCTIONS_DECL__ float expm1(float a); __MATH_FUNCTIONS_DECL__ float log2(float a); __MATH_FUNCTIONS_DECL__ float log1p(float a); __MATH_FUNCTIONS_DECL__ float acosh(float a); __MATH_FUNCTIONS_DECL__ float asinh(float a); __MATH_FUNCTIONS_DECL__ float atanh(float a); __MATH_FUNCTIONS_DECL__ float hypot(float a, float b); __MATH_FUNCTIONS_DECL__ float cbrt(float a); __MATH_FUNCTIONS_DECL__ float erf(float a); __MATH_FUNCTIONS_DECL__ float erfc(float a); __MATH_FUNCTIONS_DECL__ float lgamma(float a); __MATH_FUNCTIONS_DECL__ float tgamma(float a); __MATH_FUNCTIONS_DECL__ float copysign(float a, float b); __MATH_FUNCTIONS_DECL__ float nextafter(float a, float b); __MATH_FUNCTIONS_DECL__ float remainder(float a, float b); __MATH_FUNCTIONS_DECL__ float remquo(float a, float b, int *quo); __MATH_FUNCTIONS_DECL__ float round(float a); __MATH_FUNCTIONS_DECL__ long int lround(float a); __MATH_FUNCTIONS_DECL__ long long int llround(float a); __MATH_FUNCTIONS_DECL__ float trunc(float a); __MATH_FUNCTIONS_DECL__ float rint(float a); __MATH_FUNCTIONS_DECL__ long int lrint(float a); __MATH_FUNCTIONS_DECL__ long long int llrint(float a); __MATH_FUNCTIONS_DECL__ float nearbyint(float a); __MATH_FUNCTIONS_DECL__ float fdim(float a, float b); __MATH_FUNCTIONS_DECL__ float fma(float a, float b, float c); __MATH_FUNCTIONS_DECL__ float fmax(float a, float b); __MATH_FUNCTIONS_DECL__ float fmin(float a, float b); #endif #endif #else extern __host__ __device__ __cudart_builtin__ float __cdecl logb(float) throw(); extern __host__ __device__ __cudart_builtin__ int __cdecl ilogb(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl scalbn(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl scalbln(float, long int) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl exp2(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl expm1(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl log2(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl log1p(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl acosh(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl asinh(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl atanh(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl hypot(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl cbrt(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl erf(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl erfc(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl lgamma(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl tgamma(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl copysign(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl nextafter(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl remainder(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl remquo(float, float, int *) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl round(float) throw(); extern __host__ __device__ __cudart_builtin__ long int __cdecl lround(float) throw(); extern __host__ __device__ __cudart_builtin__ long long int __cdecl llround(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl trunc(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl rint(float) throw(); extern __host__ __device__ __cudart_builtin__ long int __cdecl lrint(float) throw(); extern __host__ __device__ __cudart_builtin__ long long int __cdecl llrint(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl nearbyint(float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl fdim(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl fma(float, float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl fmax(float, float) throw(); extern __host__ __device__ __cudart_builtin__ float __cdecl fmin(float, float) throw(); #endif __MATH_FUNCTIONS_DECL__ float exp10(const float a); __MATH_FUNCTIONS_DECL__ float rsqrt(const float a); __MATH_FUNCTIONS_DECL__ float rcbrt(const float a); __MATH_FUNCTIONS_DECL__ float sinpi(const float a); __MATH_FUNCTIONS_DECL__ float cospi(const float a); __MATH_FUNCTIONS_DECL__ void sincospi(const float a, float *const sptr, float *const cptr); __MATH_FUNCTIONS_DECL__ void sincos(const float a, float *const sptr, float *const cptr); __MATH_FUNCTIONS_DECL__ float j0(const float a); __MATH_FUNCTIONS_DECL__ float j1(const float a); __MATH_FUNCTIONS_DECL__ float jn(const int n, const float a); __MATH_FUNCTIONS_DECL__ float y0(const float a); __MATH_FUNCTIONS_DECL__ float y1(const float a); __MATH_FUNCTIONS_DECL__ float yn(const int n, const float a); __MATH_FUNCTIONS_DEVICE_DECL__ float cyl_bessel_i0(const float a); __MATH_FUNCTIONS_DEVICE_DECL__ float cyl_bessel_i1(const float a); __MATH_FUNCTIONS_DECL__ float erfinv(const float a); __MATH_FUNCTIONS_DECL__ float erfcinv(const float a); __MATH_FUNCTIONS_DECL__ float normcdfinv(const float a); __MATH_FUNCTIONS_DECL__ float normcdf(const float a); __MATH_FUNCTIONS_DECL__ float erfcx(const float a); __MATH_FUNCTIONS_DECL__ double copysign(const double a, const float b); __MATH_FUNCTIONS_DECL__ double copysign(const float a, const double b); __MATH_FUNCTIONS_DECL__ unsigned int min(const unsigned int a, const unsigned int b); __MATH_FUNCTIONS_DECL__ unsigned int min(const int a, const unsigned int b); __MATH_FUNCTIONS_DECL__ unsigned int min(const unsigned int a, const int b); __MATH_FUNCTIONS_DECL__ long int min(const long int a, const long int b); __MATH_FUNCTIONS_DECL__ unsigned long int min(const unsigned long int a, const unsigned long int b); __MATH_FUNCTIONS_DECL__ unsigned long int min(const long int a, const unsigned long int b); __MATH_FUNCTIONS_DECL__ unsigned long int min(const unsigned long int a, const long int b); __MATH_FUNCTIONS_DECL__ long long int min(const long long int a, const long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int min(const unsigned long long int a, const unsigned long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int min(const long long int a, const unsigned long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int min(const unsigned long long int a, const long long int b); __MATH_FUNCTIONS_DECL__ float min(const float a, const float b); __MATH_FUNCTIONS_DECL__ double min(const double a, const double b); __MATH_FUNCTIONS_DECL__ double min(const float a, const double b); __MATH_FUNCTIONS_DECL__ double min(const double a, const float b); __MATH_FUNCTIONS_DECL__ unsigned int max(const unsigned int a, const unsigned int b); __MATH_FUNCTIONS_DECL__ unsigned int max(const int a, const unsigned int b); __MATH_FUNCTIONS_DECL__ unsigned int max(const unsigned int a, const int b); __MATH_FUNCTIONS_DECL__ long int max(const long int a, const long int b); __MATH_FUNCTIONS_DECL__ unsigned long int max(const unsigned long int a, const unsigned long int b); __MATH_FUNCTIONS_DECL__ unsigned long int max(const long int a, const unsigned long int b); __MATH_FUNCTIONS_DECL__ unsigned long int max(const unsigned long int a, const long int b); __MATH_FUNCTIONS_DECL__ long long int max(const long long int a, const long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int max(const unsigned long long int a, const unsigned long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int max(const long long int a, const unsigned long long int b); __MATH_FUNCTIONS_DECL__ unsigned long long int max(const unsigned long long int a, const long long int b); __MATH_FUNCTIONS_DECL__ float max(const float a, const float b); __MATH_FUNCTIONS_DECL__ double max(const double a, const double b); __MATH_FUNCTIONS_DECL__ double max(const float a, const double b); __MATH_FUNCTIONS_DECL__ double max(const double a, const float b); #undef __MATH_FUNCTIONS_DECL__ #undef __MATH_FUNCTIONS_DEVICE_DECL__ extern "C"{ inline __device__ void *__nv_aligned_device_malloc(size_t size, size_t align) { __device__ void *__nv_aligned_device_malloc_impl(size_t, size_t); return __nv_aligned_device_malloc_impl(size, align); } } #endif #if !defined(__CUDACC__) #include <crt/func_macro.h> #if defined(_WIN32) #pragma warning(disable : 4211) #endif __func__(double rsqrt(double a)); __func__(double rcbrt(double a)); __func__(double sinpi(double a)); __func__(double cospi(double a)); __func__(void sincospi(double a, double *sptr, double *cptr)); __func__(double erfinv(double a)); __func__(double erfcinv(double a)); __func__(double normcdfinv(double a)); __func__(double normcdf(double a)); __func__(double erfcx(double a)); __func__(float rsqrtf(float a)); __func__(float rcbrtf(float a)); __func__(float sinpif(float a)); __func__(float cospif(float a)); __func__(void sincospif(float a, float *sptr, float *cptr)); __func__(float erfinvf(float a)); __func__(float erfcinvf(float a)); __func__(float normcdfinvf(float a)); __func__(float normcdff(float a)); __func__(float erfcxf(float a)); __func__(int min(int a, int b)); __func__(unsigned int umin(unsigned int a, unsigned int b)); __func__(long long int llmin(long long int a, long long int b)); __func__(unsigned long long int ullmin(unsigned long long int a, unsigned long long int b)); __func__(int max(int a, int b)); __func__(unsigned int umax(unsigned int a, unsigned int b)); __func__(long long int llmax(long long int a, long long int b)); __func__(unsigned long long int ullmax(unsigned long long int a, unsigned long long int b)); #if defined(_WIN32) || defined(__APPLE__) || defined (__ANDROID__) __func__(int __isnan(double a)); #endif #if defined(_WIN32) || defined(__APPLE__) || defined (__QNX__) __func__(void sincos(double a, double *sptr, double *cptr)); #endif #if defined(_WIN32) || defined(__APPLE__) __func__(double exp10(double a)); __func__(float exp10f(float a)); __func__(void sincosf(float a, float *sptr, float *cptr)); __func__(int __isinf(double a)); #endif #if (defined(_WIN32) && (!defined(_MSC_VER) || _MSC_VER < 1800)) || defined (__ANDROID__) __func__(double log2(double a)); #endif #if defined(_WIN32) __func__(int __signbit(double a)); __func__(int __finite(double a)); __func__(int __signbitl(long double a)); __func__(int __signbitf(float a)); __func__(int __finitel(long double a)); __func__(int __finitef(float a)); __func__(int __isinfl(long double a)); __func__(int __isinff(float a)); __func__(int __isnanl(long double a)); __func__(int __isnanf(float a)); #endif #if defined(_WIN32) && (!defined(_MSC_VER) || _MSC_VER < 1800) __func__(double copysign(double a, double b)); __func__(double fmax(double a, double b)); __func__(double fmin(double a, double b)); __func__(double trunc(double a)); __func__(double round(double a)); __func__(long int lround(double a)); __func__(long long int llround(double a)); __func__(double rint(double a)); __func__(double nearbyint(double a)); __func__(long int lrint(double a)); __func__(long long int llrint(double a)); __func__(double fdim(double a, double b)); __func__(double scalbn(double a, int b)); __func__(double scalbln(double a, long int b)); __func__(double exp2(double a)); __func__(double log1p(double a)); __func__(double expm1(double a)); __func__(double cbrt(double a)); __func__(double acosh(double a)); __func__(double asinh(double a)); __func__(double atanh(double a)); __func__(int ilogb(double a)); __func__(double logb(double a)); __func__(double remquo(double a, double b, int *quo)); __func__(double remainder(double a, double b)); __func__(double fma (double a, double b, double c)); __func__(double nextafter(double a, double b)); __func__(double erf(double a)); __func__(double erfc(double a)); __func__(double lgamma(double a)); __func__(unsigned long long int __internal_host_nan_kernel(const char *s)); __func__(double nan(const char *tagp)); __func__(double __host_tgamma_kernel(double a)); __func__(double __host_stirling_poly(double a)); __func__(double __host_tgamma_stirling(double a)); __func__(double tgamma(double a)); __func__(float fmaxf(float a, float b)); __func__(float fminf(float a, float b)); __func__(float roundf(float a)); __func__(long int lroundf(float a)); __func__(long long int llroundf(float a)); __func__(float truncf(float a)); __func__(float rintf(float a)); __func__(float nearbyintf(float a)); __func__(long int lrintf(float a)); __func__(long long int llrintf(float a)); __func__(float logbf(float a)); __func__(float scalblnf(float a, long int b)); __func__(float log2f(float a)); __func__(float exp2f(float a)); __func__(float acoshf(float a)); __func__(float asinhf(float a)); __func__(float atanhf(float a)); __func__(float cbrtf(float a)); __func__(float expm1f(float a)); __func__(float fdimf(float a, float b)); __func__(float log1pf(float a)); __func__(float scalbnf(float a, int b)); __func__(float fmaf(float a, float b, float c)); __func__(int ilogbf(float a)); __func__(float erff(float a)); __func__(float erfcf(float a)); __func__(float lgammaf(float a)); __func__(float tgammaf(float a)); __func__(float remquof(float a, float b, int *quo)); __func__(float remainderf(float a, float b)); __func__(float copysignf(float a, float b)); __func__(float nextafterf(float a, float b)); __func__(float nanf(const char *tagp)); #endif #if defined(_WIN32) #pragma warning(default: 4211) #endif #endif #if !defined(__CUDACC_RTC__) #include "math_functions.hpp" #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_MATH_FUNCTIONS_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_MATH_FUNCTIONS_H__ #endif #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_COMMON_FUNCTIONS_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_COMMON_FUNCTIONS_H__ #endif #if !defined(__DEVICE_LAUNCH_PARAMETERS_H__) #define __DEVICE_LAUNCH_PARAMETERS_H__ #if !defined(__VECTOR_TYPES_H__) #define __VECTOR_TYPES_H__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #ifndef __DOXYGEN_ONLY__ #if !defined(__CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__) #if defined(_MSC_VER) #pragma message("crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead.") #else #warning "crt/host_defines.h is an internal header file and must not be used directly. Please use cuda_runtime_api.h or cuda_runtime.h instead." #endif #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #define __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #if !defined(__HOST_DEFINES_H__) #define __HOST_DEFINES_H__ #if defined(__GNUC__) || (defined(__PGIC__) && defined(__linux__)) || defined(__CUDA_LIBDEVICE__) || defined(__CUDACC_RTC__) #if defined(__CUDACC_RTC__) #define __volatile__ volatile #endif #define __no_return__ \ __attribute__((noreturn)) #if defined(__CUDACC__) || defined(__CUDA_ARCH__) || defined(__CUDA_LIBDEVICE__) #define __noinline__ \ __attribute__((noinline)) #endif #define __forceinline__ \ __inline__ __attribute__((always_inline)) #define __align__(n) \ __attribute__((aligned(n))) #define __thread__ \ __thread #define __import__ #define __export__ #define __cdecl #define __annotate__(a) \ __attribute__((a)) #define __location__(a) \ __annotate__(a) #define CUDARTAPI #define CUDARTAPI_CDECL #elif defined(_MSC_VER) #if _MSC_VER >= 1400 #define __restrict__ \ __restrict #else #define __restrict__ #endif #define __inline__ \ __inline #define __no_return__ \ __declspec(noreturn) #define __noinline__ \ __declspec(noinline) #define __forceinline__ \ __forceinline #define __align__(n) \ __declspec(align(n)) #define __thread__ \ __declspec(thread) #define __import__ \ __declspec(dllimport) #define __export__ \ __declspec(dllexport) #define __annotate__(a) \ __declspec(a) #define __location__(a) \ __annotate__(__##a##__) #define CUDARTAPI \ __stdcall #define CUDARTAPI_CDECL \ __cdecl #else #define __inline__ #if !defined(__align__) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for '__align__' !!! --- #endif #if !defined(CUDARTAPI) #error --- !!! UNKNOWN COMPILER: please provide a CUDA compatible definition for 'CUDARTAPI' !!! --- #endif #endif #if (defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3 && !defined(__clang__)))) || \ (defined(_MSC_VER) && _MSC_VER < 1900) || \ (!defined(__GNUC__) && !defined(_MSC_VER)) #define __specialization_static \ static #else #define __specialization_static #endif #if !defined(__CUDACC__) && !defined(__CUDA_LIBDEVICE__) #undef __annotate__ #define __annotate__(a) #else #define __launch_bounds__(...) \ __annotate__(launch_bounds(__VA_ARGS__)) #endif #if defined(__CUDACC__) || defined(__CUDA_LIBDEVICE__) || \ defined(__GNUC__) || defined(_WIN64) #define __builtin_align__(a) \ __align__(a) #else #define __builtin_align__(a) #endif #if defined(__CUDACC__) || !defined(__grid_constant__) #define __grid_constant__ \ __location__(grid_constant) #endif #if defined(__CUDACC__) || !defined(__host__) #define __host__ \ __location__(host) #endif #if defined(__CUDACC__) || !defined(__device__) #define __device__ \ __location__(device) #endif #if defined(__CUDACC__) || !defined(__global__) #define __global__ \ __location__(global) #endif #if defined(__CUDACC__) || !defined(__shared__) #define __shared__ \ __location__(shared) #endif #if defined(__CUDACC__) || !defined(__constant__) #define __constant__ \ __location__(constant) #endif #if defined(__CUDACC__) || !defined(__managed__) #define __managed__ \ __location__(managed) #endif #if !defined(__CUDACC__) #define __device_builtin__ #define __device_builtin_texture_type__ #define __device_builtin_surface_type__ #define __cudart_builtin__ #else #define __device_builtin__ \ __location__(device_builtin) #define __device_builtin_texture_type__ \ __location__(device_builtin_texture_type) #define __device_builtin_surface_type__ \ __location__(device_builtin_surface_type) #define __cudart_builtin__ \ __location__(cudart_builtin) #endif #if defined(__CUDACC__) || !defined(__cluster_dims__) #if defined(_MSC_VER) #define __cluster_dims__(...) \ __declspec(__cluster_dims__(__VA_ARGS__)) #else #define __cluster_dims__(...) \ __attribute__((cluster_dims(__VA_ARGS__))) #endif #endif #define __CUDA_ARCH_HAS_FEATURE__(_FEAT) __CUDA_ARCH_FEAT_##_FEAT #endif #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_HOST_DEFINES_H__ #endif #endif #if !defined(__CUDACC__) && !defined(__CUDACC_RTC__) && \ defined(_WIN32) && !defined(_WIN64) #pragma warning(push) #pragma warning(disable: 4201 4408) #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ tag \ { \ union \ { \ struct { members }; \ struct { long long int :1,:0; }; \ }; \ } #else #define __cuda_builtin_vector_align8(tag, members) \ struct __device_builtin__ __align__(8) tag \ { \ members \ } #endif struct __device_builtin__ char1 { signed char x; }; struct __device_builtin__ uchar1 { unsigned char x; }; struct __device_builtin__ __align__(2) char2 { signed char x, y; }; struct __device_builtin__ __align__(2) uchar2 { unsigned char x, y; }; struct __device_builtin__ char3 { signed char x, y, z; }; struct __device_builtin__ uchar3 { unsigned char x, y, z; }; struct __device_builtin__ __align__(4) char4 { signed char x, y, z, w; }; struct __device_builtin__ __align__(4) uchar4 { unsigned char x, y, z, w; }; struct __device_builtin__ short1 { short x; }; struct __device_builtin__ ushort1 { unsigned short x; }; struct __device_builtin__ __align__(4) short2 { short x, y; }; struct __device_builtin__ __align__(4) ushort2 { unsigned short x, y; }; struct __device_builtin__ short3 { short x, y, z; }; struct __device_builtin__ ushort3 { unsigned short x, y, z; }; __cuda_builtin_vector_align8(short4, short x; short y; short z; short w;); __cuda_builtin_vector_align8(ushort4, unsigned short x; unsigned short y; unsigned short z; unsigned short w;); struct __device_builtin__ int1 { int x; }; struct __device_builtin__ uint1 { unsigned int x; }; __cuda_builtin_vector_align8(int2, int x; int y;); __cuda_builtin_vector_align8(uint2, unsigned int x; unsigned int y;); struct __device_builtin__ int3 { int x, y, z; }; struct __device_builtin__ uint3 { unsigned int x, y, z; }; struct __device_builtin__ __builtin_align__(16) int4 { int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) uint4 { unsigned int x, y, z, w; }; struct __device_builtin__ long1 { long int x; }; struct __device_builtin__ ulong1 { unsigned long x; }; #if defined(_WIN32) __cuda_builtin_vector_align8(long2, long int x; long int y;); __cuda_builtin_vector_align8(ulong2, unsigned long int x; unsigned long int y;); #else struct __device_builtin__ __align__(2*sizeof(long int)) long2 { long int x, y; }; struct __device_builtin__ __align__(2*sizeof(unsigned long int)) ulong2 { unsigned long int x, y; }; #endif struct __device_builtin__ long3 { long int x, y, z; }; struct __device_builtin__ ulong3 { unsigned long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) long4 { long int x, y, z, w; }; struct __device_builtin__ __builtin_align__(16) ulong4 { unsigned long int x, y, z, w; }; struct __device_builtin__ float1 { float x; }; #if !defined(__CUDACC__) && defined(__arm__) && \ defined(__ARM_PCS_VFP) && __GNUC__ == 4 && __GNUC_MINOR__ == 6 #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-pedantic" struct __device_builtin__ __attribute__((aligned(8))) float2 { float x; float y; float __cuda_gnu_arm_ice_workaround[0]; }; #pragma GCC poison __cuda_gnu_arm_ice_workaround #pragma GCC diagnostic pop #else __cuda_builtin_vector_align8(float2, float x; float y;); #endif struct __device_builtin__ float3 { float x, y, z; }; struct __device_builtin__ __builtin_align__(16) float4 { float x, y, z, w; }; struct __device_builtin__ longlong1 { long long int x; }; struct __device_builtin__ ulonglong1 { unsigned long long int x; }; struct __device_builtin__ __builtin_align__(16) longlong2 { long long int x, y; }; struct __device_builtin__ __builtin_align__(16) ulonglong2 { unsigned long long int x, y; }; struct __device_builtin__ longlong3 { long long int x, y, z; }; struct __device_builtin__ ulonglong3 { unsigned long long int x, y, z; }; struct __device_builtin__ __builtin_align__(16) longlong4 { long long int x, y, z ,w; }; struct __device_builtin__ __builtin_align__(16) ulonglong4 { unsigned long long int x, y, z, w; }; struct __device_builtin__ double1 { double x; }; struct __device_builtin__ __builtin_align__(16) double2 { double x, y; }; struct __device_builtin__ double3 { double x, y, z; }; struct __device_builtin__ __builtin_align__(16) double4 { double x, y, z, w; }; #if !defined(__CUDACC__) && defined(_WIN32) && !defined(_WIN64) #pragma warning(pop) #endif typedef __device_builtin__ struct char1 char1; typedef __device_builtin__ struct uchar1 uchar1; typedef __device_builtin__ struct char2 char2; typedef __device_builtin__ struct uchar2 uchar2; typedef __device_builtin__ struct char3 char3; typedef __device_builtin__ struct uchar3 uchar3; typedef __device_builtin__ struct char4 char4; typedef __device_builtin__ struct uchar4 uchar4; typedef __device_builtin__ struct short1 short1; typedef __device_builtin__ struct ushort1 ushort1; typedef __device_builtin__ struct short2 short2; typedef __device_builtin__ struct ushort2 ushort2; typedef __device_builtin__ struct short3 short3; typedef __device_builtin__ struct ushort3 ushort3; typedef __device_builtin__ struct short4 short4; typedef __device_builtin__ struct ushort4 ushort4; typedef __device_builtin__ struct int1 int1; typedef __device_builtin__ struct uint1 uint1; typedef __device_builtin__ struct int2 int2; typedef __device_builtin__ struct uint2 uint2; typedef __device_builtin__ struct int3 int3; typedef __device_builtin__ struct uint3 uint3; typedef __device_builtin__ struct int4 int4; typedef __device_builtin__ struct uint4 uint4; typedef __device_builtin__ struct long1 long1; typedef __device_builtin__ struct ulong1 ulong1; typedef __device_builtin__ struct long2 long2; typedef __device_builtin__ struct ulong2 ulong2; typedef __device_builtin__ struct long3 long3; typedef __device_builtin__ struct ulong3 ulong3; typedef __device_builtin__ struct long4 long4; typedef __device_builtin__ struct ulong4 ulong4; typedef __device_builtin__ struct float1 float1; typedef __device_builtin__ struct float2 float2; typedef __device_builtin__ struct float3 float3; typedef __device_builtin__ struct float4 float4; typedef __device_builtin__ struct longlong1 longlong1; typedef __device_builtin__ struct ulonglong1 ulonglong1; typedef __device_builtin__ struct longlong2 longlong2; typedef __device_builtin__ struct ulonglong2 ulonglong2; typedef __device_builtin__ struct longlong3 longlong3; typedef __device_builtin__ struct ulonglong3 ulonglong3; typedef __device_builtin__ struct longlong4 longlong4; typedef __device_builtin__ struct ulonglong4 ulonglong4; typedef __device_builtin__ struct double1 double1; typedef __device_builtin__ struct double2 double2; typedef __device_builtin__ struct double3 double3; typedef __device_builtin__ struct double4 double4; struct __device_builtin__ dim3 { unsigned int x, y, z; #if defined(__cplusplus) #if __cplusplus >= 201103L __host__ __device__ constexpr dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ constexpr dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ constexpr operator uint3(void) const { return uint3{x, y, z}; } #else __host__ __device__ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) {} __host__ __device__ dim3(uint3 v) : x(v.x), y(v.y), z(v.z) {} __host__ __device__ operator uint3(void) const { uint3 t; t.x = x; t.y = y; t.z = z; return t; } #endif #endif }; typedef __device_builtin__ struct dim3 dim3; #undef __cuda_builtin_vector_align8 #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_VECTOR_TYPES_H__ #endif #endif #if !defined(__STORAGE__) #if defined(__CUDACC_RTC__) #define __STORAGE__ \ extern const __device__ #else #define __STORAGE__ \ extern const #endif #endif #if defined(__cplusplus) extern "C" { #endif uint3 __device_builtin__ __STORAGE__ threadIdx; uint3 __device_builtin__ __STORAGE__ blockIdx; dim3 __device_builtin__ __STORAGE__ blockDim; dim3 __device_builtin__ __STORAGE__ gridDim; int __device_builtin__ __STORAGE__ warpSize; #undef __STORAGE__ #if defined(__cplusplus) } #endif #if !defined(__cudaGet_threadIdx) #define __cudaGet_threadIdx() \ threadIdx #endif #if !defined(__cudaGet_blockIdx) #define __cudaGet_blockIdx() \ blockIdx #endif #if !defined(__cudaGet_blockDim) #define __cudaGet_blockDim() \ blockDim #endif #if !defined(__cudaGet_gridDim) #define __cudaGet_gridDim() \ gridDim #endif #if !defined(__cudaGet_warpSize) #define __cudaGet_warpSize() \ warpSize #endif #endif #else #endif #endif #if defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) #define __CUDA_DEPRECATED #elif defined(_MSC_VER) #define __CUDA_DEPRECATED __declspec(deprecated) #elif defined(__GNUC__) #define __CUDA_DEPRECATED __attribute__((deprecated)) #else #define __CUDA_DEPRECATED #endif #if defined(__cplusplus) && !defined(__CUDACC_RTC__) #if __cplusplus >= 201103 #include <utility> #endif template<class T> static __inline__ __host__ cudaError_t cudaLaunchKernel( const T *func, dim3 gridDim, dim3 blockDim, void **args, size_t sharedMem = 0, cudaStream_t stream = 0 ) { return ::cudaLaunchKernel((const void *)func, gridDim, blockDim, args, sharedMem, stream); } #if __cplusplus >= 201103 || defined(__DOXYGEN_ONLY__) template<typename... ExpTypes, typename... ActTypes> static __inline__ __host__ cudaError_t cudaLaunchKernelEx( const cudaLaunchConfig_t *config, void (*kernel)(ExpTypes...), ActTypes &&... args ) { return [&](ExpTypes... coercedArgs){ void *pArgs[] = { &coercedArgs... }; return ::cudaLaunchKernelExC(config, (const void *)kernel, pArgs); }(std::forward<ActTypes>(args)...); } #endif template<class T> static __inline__ __host__ cudaError_t cudaLaunchCooperativeKernel( const T *func, dim3 gridDim, dim3 blockDim, void **args, size_t sharedMem = 0, cudaStream_t stream = 0 ) { return ::cudaLaunchCooperativeKernel((const void *)func, gridDim, blockDim, args, sharedMem, stream); } static __inline__ __host__ cudaError_t cudaEventCreate( cudaEvent_t *event, unsigned int flags ) { return ::cudaEventCreateWithFlags(event, flags); } static __inline__ __host__ cudaError_t cudaGraphInstantiate( cudaGraphExec_t *pGraphExec, cudaGraph_t graph, cudaGraphNode_t *pErrorNode, char *pLogBuffer, size_t bufferSize ) { (void)pErrorNode; (void)pLogBuffer; (void)bufferSize; return ::cudaGraphInstantiate(pGraphExec, graph, 0); } static __inline__ __host__ cudaError_t cudaMallocHost( void **ptr, size_t size, unsigned int flags ) { return ::cudaHostAlloc(ptr, size, flags); } template<class T> static __inline__ __host__ cudaError_t cudaHostAlloc( T **ptr, size_t size, unsigned int flags ) { return ::cudaHostAlloc((void**)(void*)ptr, size, flags); } template<class T> static __inline__ __host__ cudaError_t cudaHostGetDevicePointer( T **pDevice, void *pHost, unsigned int flags ) { return ::cudaHostGetDevicePointer((void**)(void*)pDevice, pHost, flags); } template<class T> static __inline__ __host__ cudaError_t cudaMallocManaged( T **devPtr, size_t size, unsigned int flags = cudaMemAttachGlobal ) { return ::cudaMallocManaged((void**)(void*)devPtr, size, flags); } template<class T> static __inline__ __host__ cudaError_t cudaStreamAttachMemAsync( cudaStream_t stream, T *devPtr, size_t length = 0, unsigned int flags = cudaMemAttachSingle ) { return ::cudaStreamAttachMemAsync(stream, (void*)devPtr, length, flags); } template<class T> static __inline__ __host__ cudaError_t cudaMalloc( T **devPtr, size_t size ) { return ::cudaMalloc((void**)(void*)devPtr, size); } template<class T> static __inline__ __host__ cudaError_t cudaMallocHost( T **ptr, size_t size, unsigned int flags = 0 ) { return cudaMallocHost((void**)(void*)ptr, size, flags); } template<class T> static __inline__ __host__ cudaError_t cudaMallocPitch( T **devPtr, size_t *pitch, size_t width, size_t height ) { return ::cudaMallocPitch((void**)(void*)devPtr, pitch, width, height); } static __inline__ __host__ cudaError_t cudaMallocAsync( void **ptr, size_t size, cudaMemPool_t memPool, cudaStream_t stream ) { return ::cudaMallocFromPoolAsync(ptr, size, memPool, stream); } template<class T> static __inline__ __host__ cudaError_t cudaMallocAsync( T **ptr, size_t size, cudaMemPool_t memPool, cudaStream_t stream ) { return ::cudaMallocFromPoolAsync((void**)(void*)ptr, size, memPool, stream); } template<class T> static __inline__ __host__ cudaError_t cudaMallocAsync( T **ptr, size_t size, cudaStream_t stream ) { return ::cudaMallocAsync((void**)(void*)ptr, size, stream); } template<class T> static __inline__ __host__ cudaError_t cudaMallocFromPoolAsync( T **ptr, size_t size, cudaMemPool_t memPool, cudaStream_t stream ) { return ::cudaMallocFromPoolAsync((void**)(void*)ptr, size, memPool, stream); } #if defined(__CUDACC__) template<class T> static __inline__ __host__ cudaError_t cudaMemcpyToSymbol( const T &symbol, const void *src, size_t count, size_t offset = 0, enum cudaMemcpyKind kind = cudaMemcpyHostToDevice ) { return ::cudaMemcpyToSymbol((const void*)&symbol, src, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaMemcpyToSymbolAsync( const T &symbol, const void *src, size_t count, size_t offset = 0, enum cudaMemcpyKind kind = cudaMemcpyHostToDevice, cudaStream_t stream = 0 ) { return ::cudaMemcpyToSymbolAsync((const void*)&symbol, src, count, offset, kind, stream); } template<class T> static __inline__ __host__ cudaError_t cudaMemcpyFromSymbol( void *dst, const T &symbol, size_t count, size_t offset = 0, enum cudaMemcpyKind kind = cudaMemcpyDeviceToHost ) { return ::cudaMemcpyFromSymbol(dst, (const void*)&symbol, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaMemcpyFromSymbolAsync( void *dst, const T &symbol, size_t count, size_t offset = 0, enum cudaMemcpyKind kind = cudaMemcpyDeviceToHost, cudaStream_t stream = 0 ) { return ::cudaMemcpyFromSymbolAsync(dst, (const void*)&symbol, count, offset, kind, stream); } template<class T> static __inline__ __host__ cudaError_t cudaGraphAddMemcpyNodeToSymbol( cudaGraphNode_t *pGraphNode, cudaGraph_t graph, const cudaGraphNode_t *pDependencies, size_t numDependencies, const T &symbol, const void* src, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphAddMemcpyNodeToSymbol(pGraphNode, graph, pDependencies, numDependencies, (const void*)&symbol, src, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaGraphAddMemcpyNodeFromSymbol( cudaGraphNode_t* pGraphNode, cudaGraph_t graph, const cudaGraphNode_t* pDependencies, size_t numDependencies, void* dst, const T &symbol, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphAddMemcpyNodeFromSymbol(pGraphNode, graph, pDependencies, numDependencies, dst, (const void*)&symbol, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaGraphMemcpyNodeSetParamsToSymbol( cudaGraphNode_t node, const T &symbol, const void* src, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphMemcpyNodeSetParamsToSymbol(node, (const void*)&symbol, src, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaGraphMemcpyNodeSetParamsFromSymbol( cudaGraphNode_t node, void* dst, const T &symbol, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphMemcpyNodeSetParamsFromSymbol(node, dst, (const void*)&symbol, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaGraphExecMemcpyNodeSetParamsToSymbol( cudaGraphExec_t hGraphExec, cudaGraphNode_t node, const T &symbol, const void* src, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphExecMemcpyNodeSetParamsToSymbol(hGraphExec, node, (const void*)&symbol, src, count, offset, kind); } template<class T> static __inline__ __host__ cudaError_t cudaGraphExecMemcpyNodeSetParamsFromSymbol( cudaGraphExec_t hGraphExec, cudaGraphNode_t node, void* dst, const T &symbol, size_t count, size_t offset, enum cudaMemcpyKind kind) { return ::cudaGraphExecMemcpyNodeSetParamsFromSymbol(hGraphExec, node, dst, (const void*)&symbol, count, offset, kind); } static __inline__ __host__ cudaError_t CUDARTAPI cudaGraphExecUpdate(cudaGraphExec_t hGraphExec, cudaGraph_t hGraph, cudaGraphNode_t *hErrorNode_out, enum cudaGraphExecUpdateResult *updateResult_out) { cudaGraphExecUpdateResultInfo resultInfo; cudaError_t status = cudaGraphExecUpdate(hGraphExec, hGraph, &resultInfo); if (hErrorNode_out) { *hErrorNode_out = resultInfo.errorNode; } if (updateResult_out) { *updateResult_out = resultInfo.result; } return status; } #if __cplusplus >= 201103 template<class T> static __inline__ __host__ cudaError_t cudaUserObjectCreate( cudaUserObject_t *object_out, T *objectToWrap, unsigned int initialRefcount, unsigned int flags) { return ::cudaUserObjectCreate( object_out, objectToWrap, [](void *vpObj) { delete reinterpret_cast<T *>(vpObj); }, initialRefcount, flags); } template<class T> static __inline__ __host__ cudaError_t cudaUserObjectCreate( cudaUserObject_t *object_out, T *objectToWrap, unsigned int initialRefcount, cudaUserObjectFlags flags) { return cudaUserObjectCreate(object_out, objectToWrap, initialRefcount, (unsigned int)flags); } #endif template<class T> static __inline__ __host__ cudaError_t cudaGetSymbolAddress( void **devPtr, const T &symbol ) { return ::cudaGetSymbolAddress(devPtr, (const void*)&symbol); } template<class T> static __inline__ __host__ cudaError_t cudaGetSymbolSize( size_t *size, const T &symbol ) { return ::cudaGetSymbolSize(size, (const void*)&symbol); } template<class T> static __inline__ __host__ cudaError_t cudaFuncSetCacheConfig( T *func, enum cudaFuncCache cacheConfig ) { return ::cudaFuncSetCacheConfig((const void*)func, cacheConfig); } template<class T> static __inline__ __host__ cudaError_t cudaFuncSetSharedMemConfig( T *func, enum cudaSharedMemConfig config ) { return ::cudaFuncSetSharedMemConfig((const void*)func, config); } #endif template<class T> static __inline__ __host__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessor( int *numBlocks, T func, int blockSize, size_t dynamicSMemSize) { return ::cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, (const void*)func, blockSize, dynamicSMemSize, cudaOccupancyDefault); } template<class T> static __inline__ __host__ cudaError_t cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( int *numBlocks, T func, int blockSize, size_t dynamicSMemSize, unsigned int flags) { return ::cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, (const void*)func, blockSize, dynamicSMemSize, flags); } class __cudaOccupancyB2DHelper { size_t n; public: inline __host__ CUDART_DEVICE __cudaOccupancyB2DHelper(size_t n_) : n(n_) {} inline __host__ CUDART_DEVICE size_t operator()(int) { return n; } }; template<typename UnaryFunction, class T> static __inline__ __host__ CUDART_DEVICE cudaError_t cudaOccupancyMaxPotentialBlockSizeVariableSMemWithFlags( int *minGridSize, int *blockSize, T func, UnaryFunction blockSizeToDynamicSMemSize, int blockSizeLimit = 0, unsigned int flags = 0) { cudaError_t status; int device; struct cudaFuncAttributes attr; int maxThreadsPerMultiProcessor; int warpSize; int devMaxThreadsPerBlock; int multiProcessorCount; int funcMaxThreadsPerBlock; int occupancyLimit; int granularity; int maxBlockSize = 0; int numBlocks = 0; int maxOccupancy = 0; int blockSizeToTryAligned; int blockSizeToTry; int blockSizeLimitAligned; int occupancyInBlocks; int occupancyInThreads; size_t dynamicSMemSize; if (!minGridSize || !blockSize || !func) { return cudaErrorInvalidValue; } status = ::cudaGetDevice(&device); if (status != cudaSuccess) { return status; } status = cudaDeviceGetAttribute( &maxThreadsPerMultiProcessor, cudaDevAttrMaxThreadsPerMultiProcessor, device); if (status != cudaSuccess) { return status; } status = cudaDeviceGetAttribute( &warpSize, cudaDevAttrWarpSize, device); if (status != cudaSuccess) { return status; } status = cudaDeviceGetAttribute( &devMaxThreadsPerBlock, cudaDevAttrMaxThreadsPerBlock, device); if (status != cudaSuccess) { return status; } status = cudaDeviceGetAttribute( &multiProcessorCount, cudaDevAttrMultiProcessorCount, device); if (status != cudaSuccess) { return status; } status = cudaFuncGetAttributes(&attr, func); if (status != cudaSuccess) { return status; } funcMaxThreadsPerBlock = attr.maxThreadsPerBlock; occupancyLimit = maxThreadsPerMultiProcessor; granularity = warpSize; if (blockSizeLimit == 0) { blockSizeLimit = devMaxThreadsPerBlock; } if (devMaxThreadsPerBlock < blockSizeLimit) { blockSizeLimit = devMaxThreadsPerBlock; } if (funcMaxThreadsPerBlock < blockSizeLimit) { blockSizeLimit = funcMaxThreadsPerBlock; } blockSizeLimitAligned = ((blockSizeLimit + (granularity - 1)) / granularity) * granularity; for (blockSizeToTryAligned = blockSizeLimitAligned; blockSizeToTryAligned > 0; blockSizeToTryAligned -= granularity) { if (blockSizeLimit < blockSizeToTryAligned) { blockSizeToTry = blockSizeLimit; } else { blockSizeToTry = blockSizeToTryAligned; } dynamicSMemSize = blockSizeToDynamicSMemSize(blockSizeToTry); status = cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( &occupancyInBlocks, func, blockSizeToTry, dynamicSMemSize, flags); if (status != cudaSuccess) { return status; } occupancyInThreads = blockSizeToTry * occupancyInBlocks; if (occupancyInThreads > maxOccupancy) { maxBlockSize = blockSizeToTry; numBlocks = occupancyInBlocks; maxOccupancy = occupancyInThreads; } if (occupancyLimit == maxOccupancy) { break; } } *minGridSize = numBlocks * multiProcessorCount; *blockSize = maxBlockSize; return status; } template<typename UnaryFunction, class T> static __inline__ __host__ CUDART_DEVICE cudaError_t cudaOccupancyMaxPotentialBlockSizeVariableSMem( int *minGridSize, int *blockSize, T func, UnaryFunction blockSizeToDynamicSMemSize, int blockSizeLimit = 0) { return cudaOccupancyMaxPotentialBlockSizeVariableSMemWithFlags(minGridSize, blockSize, func, blockSizeToDynamicSMemSize, blockSizeLimit, cudaOccupancyDefault); } template<class T> static __inline__ __host__ CUDART_DEVICE cudaError_t cudaOccupancyMaxPotentialBlockSize( int *minGridSize, int *blockSize, T func, size_t dynamicSMemSize = 0, int blockSizeLimit = 0) { return cudaOccupancyMaxPotentialBlockSizeVariableSMemWithFlags(minGridSize, blockSize, func, __cudaOccupancyB2DHelper(dynamicSMemSize), blockSizeLimit, cudaOccupancyDefault); } template<class T> static __inline__ __host__ cudaError_t cudaOccupancyAvailableDynamicSMemPerBlock( size_t *dynamicSmemSize, T func, int numBlocks, int blockSize) { return ::cudaOccupancyAvailableDynamicSMemPerBlock(dynamicSmemSize, (const void*)func, numBlocks, blockSize); } template<class T> static __inline__ __host__ CUDART_DEVICE cudaError_t cudaOccupancyMaxPotentialBlockSizeWithFlags( int *minGridSize, int *blockSize, T func, size_t dynamicSMemSize = 0, int blockSizeLimit = 0, unsigned int flags = 0) { return cudaOccupancyMaxPotentialBlockSizeVariableSMemWithFlags(minGridSize, blockSize, func, __cudaOccupancyB2DHelper(dynamicSMemSize), blockSizeLimit, flags); } template<class T> static __inline__ __host__ cudaError_t cudaOccupancyMaxPotentialClusterSize( int *clusterSize, T *func, const cudaLaunchConfig_t *config) { return ::cudaOccupancyMaxPotentialClusterSize(clusterSize, (const void*)func, config); } template<class T> static __inline__ __host__ cudaError_t cudaOccupancyMaxActiveClusters( int *numClusters, T *func, const cudaLaunchConfig_t *config) { return ::cudaOccupancyMaxActiveClusters(numClusters, (const void*)func, config); } #if defined __CUDACC__ template<class T> static __inline__ __host__ cudaError_t cudaFuncGetAttributes( struct cudaFuncAttributes *attr, T *entry ) { return ::cudaFuncGetAttributes(attr, (const void*)entry); } template<class T> static __inline__ __host__ cudaError_t cudaFuncSetAttribute( T *entry, enum cudaFuncAttribute attr, int value ) { return ::cudaFuncSetAttribute((const void*)entry, attr, value); } template<class T> static __inline__ __host__ cudaError_t cudaGetKernel( cudaKernel_t *kernelPtr, const T *entryFuncAddr ) { return ::cudaGetKernel(kernelPtr, (const void *)entryFuncAddr); } #endif #endif #if !defined(__CUDACC_RTC__) #if defined(__GNUC__) #if defined(__clang__) || (!defined(__PGIC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) #pragma GCC diagnostic pop #endif #elif defined(_MSC_VER) #pragma warning(pop) #endif #endif #undef __CUDA_DEPRECATED #if defined(__UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_H__) #undef __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__ #undef __UNDEF_CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS_CUDA_RUNTIME_H__ #endif #endif BC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��U( ��EDT��*�Re�H)#�E!�-��B��S�0�� A9@S��L@W��~ʪ�@[��@�WE0� �T@c�E��,2��(BhQZG�X��"?��[�)RE�"�Ȓ"XB�E�!��/��E{E@�/��(�Q`H��F�ЈateR�L#ň�"��5F��"��IF��ʨ2��\F�Q2EHF3ڌ�):��ge��2\Q$Z�(�"�W(�V�+�*�P+�k�A�� pE\��ke��.K�2\qW($[��K��W��V�+��p+��B�� !��pEb��oe��1R�2\�X5$\�H�"�WPFW�+.��+��%Cʕ��!��`�(b�*�E��]W�ȡ�@Z��!��Z��E��;��E��:�ȼ"tw�g(�zz�zE��/j��F��(�"�g�W��ct_�{��3�/�a�ˈ?C��/��?��?#�CN�,�@� ��,�k��;� �qE\w�q��`!��b\qAX�+�2��W�Da1��,,��qaX�+��;� qEb@� ��B\��_~!�q�f��/Db!��r�D,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8no�P�� @��av@�1�7��P�av@h��Aw�B"�� F.�� @��\0��`�2�`y7��|r�8�n ��q��@;��ແv@��w�R-�� ].�� @�ڿ\0��`D�A.G� @�j� �C� ��h��1z�~5�� v@x�q��@;�x�8�n �P;��7��Y@��`��- �� @��@�� }� �qz�FPp�@��v@�(��q��@;��\�8To �P{ �`�7�hOW0��`��+G� @��]�� z��Yp〽�v@�-��q��@;�\�8do �P��`�7�h�W0��`��+G� @�:�� "��-@��r� �H�9xC�� `��%0D��!��"��Y@�`u�, �H�:`C$&��!R��@� n�� `�t�C@��w��!��0D��2 �Hq<hf@�i� �"��0h@��K �"��C��8 `��D@�v{��!|�0DR�� Ha>px@��χ+�"��C$x�Q `�T�6 @�~~�J��!��\)0D��: �H�?p�@�� `��C�Y$�`�"�$��C��$�W `��@�ȓ�f��!R�V,0D�Q� �HoJ�@��N �"�)��C$Y%�^``�d�� @�J�x��!��8P0D"[2�HtKD��t��@"�.!� C$�%$y@`��T,�Z��!R?P0Drc"��HLH"��)� �$@"�31� C${&��@`��,�z�hb��!�ZM0D�k¹ �H�M@:@�鶉�'"�7�C��& � `�t�T�r�x��!��Z0D�sr�H�NDq@�)׉�."�;!�C$~'$� `��B��!R�i0D}B: �H�OL�@��5"�?A�C$,�� `�$�Ee�r��!+o0D�� H�X\�@��%��<"�e!�C$�,&� `�T��r��!RzC|0Db�B��H�Y@�@��E�!>�"�h!�C�1-�� `��3��`��!R�H|0D��H�Z0�@�i]g>"�k��C��-�� `�Բ�3�v��!��|0D��"��H�[L�@�i|g>"�o�C�.�� `��4�~�p��!R�|0D��H�\@�@�i��i>"�sQ�C��.�� `�d�E5�ֺ`��!�]L}0D킩�H�]0�@�i��>"�w��C��.�� `�T�s�f�X��!��J~0DB�b��Hs^$�@��ϋe?"�z��C�X/�� `��E�@�̽`��!��~0D��H&_$�@��$?"�|��C$�/�� `��ű@��H��!R��~0D �"��H�_ �@�I��d?"�~q�C$�/�� `��@��8��!��E0D��H&h�@��?�"ɠq�C�4� `�ԃ��@��!R�0D�C#��H{h�@�) �?"��Q�C$W4� `�d�1"@�t�@J��!�8J�0DbG9�H�h$'@��?"��A�C$�4�� `��@�n�HF��!Xň0D*KCQ�Hoi*@��.��?"��A�C��4 `��"@�`�PJ��!�nF�0DO��H�i�@��=��?"ѧA�C$�4�� `��F�@�J�@��!R��0D�Q��H9j�@��G �?"��Q�C$55�� `�$�F�@��PF��!R�ƈ0D:U��H�j�@�IV �?"�Q�C�_5�� `�t�F�@��H��!R�F~0D�Wc)�H&k%@��e �E"嬁�C��5�� `��F�@��H��!��0D�[��H�k�@�Iu��?"ͮa�C��5�� `�$��@��H��!R��0D�_#��H l�@�Ʉ eD"��a�C$6�� `��F�@��8��!��0Drc��Hvl�@�i� �?"��q�C$V6�� `�d��@�z�hF��!�:Ɉ0D�g��H!m(#@�i� gD"崡�C��6 `��#@��H��!h�0D2mC��H�m2@�鶍EF"��C$�6�� `�4߆�@��@��!R��0D2q#��H/n2@�� EF"��q�C$37�� `�4��@��HF��!��ň0DZu#��H�n4@� ׍dD"��Q�C$s7� `��F1"@��H��!R��0Dy#��H,o�@�I��?"��a�C��7�� `��@��H��!R��0D�{��H�o$6@� � �F"��q�C��7�� `�4��@��0��!��0D��H&x8@�)�?"��1�C$<�� `��@�v�8p��!��0D*�C��Hwx8�@��?"��C�r<�� `��@��X��!RH�0DJ��H8y,�@�i,��?"��C$�<�� `��@�J�ht��!�l�0D�ͣ��H�y�@�i=$>"��C��<0 `�4�G&@�P�@��!��0D��H>z `@��LL"�遀 C$;=0 `��&�L�@F��!��0D�Ճ� �H�z >@�I\DG"��A�C$w=�� `�d��@��!��0DZ��H3{`@�)gL"�큀 C��=0 `��&��@��!R��0DJ��H�{ <@�)w�G�"��C��= `�Ծ�#&��`��! ��0D�� Hc|<b@�I��GL"�� C$Y>1 `��G$&��h��!R?�0Dr�! �H;}<b@�I�FL�"�� C��>1 `��#&��`��!{��0D�� H'~<b@��FL�"�� C�8?1 `��#&�X�`��!��0D:�� H�~8b@�)�EL�"�� C$�?1 `�T��#&��`��!��0Dz�� H�@C��?��!��@��f `�4��c&��q��!"��0D E�1 �H��@h@� M�"�#r� C�zD 6 `��ȡ&0�B)��!�I"��0D�I� �Hf� p@��-N�"�%�� C$�D09 `�d�H�#@�~Ix��!z" �0D�O$� �H��`r@�IF��G�"�(��C$2E `�ԦH�#@��!��" �0D�U$� �H��$<@�)]��G�"�+�� C$�E�� `��ȳ@��y��!R�"�~0Db]��H�@�@�ih?�"�0�C�F�� `��g'�Nq��!R-#�0Dg�� H�8x@��N"�4�� C$�F< `��ȃ'@�Ty��!Rn#��0D�o�� H#�8|@�)��O�"�8�� C�6G<= `��'@�ty��!R�#�0D�w�� H'�x~@��O"�>�@��L"�? C��G3 `�䂉c&@�x0y��!�&��0D��1�H��4f@��L"�c C�wL3 `��Ic&@�V2y��!�N&��0DB��1�Hw�8f@��4��L"�f� C$�L3 `��Ic&@��3y��!�&��0D��D�Hd� l@��MKT "�j2� C�ZM(5�`��I�&@��5��!R�&�0D��$��Hn�d�@��t�("�n�� C��M�8 `�D� F*@4�x8y��!�'��0D��3�H�|�@�)��00�Hb�x�@� ��"�v��C$�N�Z `�� W��<Y��!�'�|�0Dj�dB�H��X��MU"�|�� C$�OT `��Ʌ*@<�`?��!�'S�0D��H#�@�@�i�)>"��I��Pa��!�(*�|0DzE��H��<�@��i>"��"�C��T�� `�$�J4��Ra^��!�_*ϫ0DrMEz�H�D�@�)>&V"�� C�U�X `��+��T�b��!R�*R�0DZU%��H��D�@��]fV"�� C$�U�Y `��J4+��Vab��!��*L�0D]��H��<�@�)|)V"ͯ"� C�V�X `�d�J+��X�j��!�+׭0D�ee��H�X�@�I��V"�� C$�V�Z `��U+�n[�j��!�{+S�0D q�� H/�8�@��+W"͹�� C�UW�\ `�$��+��]�z��!�+V�0D�y��Ht�`�@��W"��r� C��W�^ `�$��+��p�~��!�.د0D��H�L�@��W"��b� C$�\�^ `�ė�+�hs�z��!{.W�0D��H:�,�@��L�%\"�鲌C�=]�q `�d��b.D�buY �H��,�@��\(]"��C��]�w `��K$/��vy��!�.�0D:݅��H��,��|��\�"��@��wY��!� /��0D��% �Hi�H$@�i��d"��C�^^�� `��΋3�FzYd��!�K/K�0D��%��Hj�D;@�i��g"��C��^�p `�T�3.D��{a��!R/��0DJ�"��C�0_�t `�$�K�.�F}��!R�/��0D:��H��@�@��j_"��C$�_s@`�D�0D�� H��H�@��)d"��2�C�d�� `�d�L�2�ΐ�Z��!2�0DzE��H��05@�� g"�##�C��d&� `��.�Ԓa��!�]2̸0DM�"�H��0�@�I7 `�d�.�䓉��!҈2�0D�QFz�Hf�H�@� O�^"�*�C$\e�} `�Ԯ3.H��a� �!R�2@�)f&\"�,#�C��e�� `��D2�h��N��!��2��0D�_&j�H)�\0@� �Ff"�1��C$<f$� `��ʌ�3�x��|��!�;3L�0D�g� �H'�0�@�i�f\�"�4ØC��f��!�]3L�0Dm&J�H��D�@��]"�7#�C�g�z `��/�Ҝ��!��3̸ 0DRu�1$�H��0C$qg�p `��/�@��B��!��3�0D*{f:�Hx�D*@�)��e"�>s�C��g� `��S3�R��p��!�6��0Dj�f��H��0�@��&\"�bÌC$rl�q `�� c.D��a �H(�0�@��&�(]"�e�C$�l�w `��$/�t��!�z6�0D�Ϧ��H/�0��iG�\�"�i�@��a��!ҟ6��0D��F �H��L$@��^��d"�l�C$�m�� `��3�D�ad��!��6L�0D��F��H��H;@��g"�pÄC�n�p `�� 3.D��a��!R(7��0Dj冁"�rÄC�rn�t `�d�M�.�N��!RN7��0DZ��H{�@�@�i��j_"�wÌC$�ns@`�� 0D"� �H0�H�@�I�)d"�y3�C�So�� `��M�2�ֽ�Z��!�7�0D��H?�05@�i� g"�}#�C��o&� `�$� 6�ܿa��!��7��00D2A�&�H2�0d��l"��À C$;t$� `��6�tљ��!R;:��0DI'��H1�Dr@�I,�n"�È C$�t��`��c6�x�a��!z:��0D�Og��H.�Pz@�IL�o"٩#� C$Tu$� `��&:��aH��!��:��0DRYgB�H=�L�@� n�u"��À C$�u� `�Ļ#6L��a�� !�:�00Da�1�H,�0`@� � m"��#� C$>v&� `��6��ى��!�<;��0D"iǢ�H:�0b�ɬ�l�"ɵØ C$�v� `��d7�zۙ��!|;��0D"q��H8�H~@�I� t"񹃉C$Zw� `��c:��ݙP��!ҿ;��0D�y�b�Hf�0`@�I�l"��È C$�w�``�d�C6L��a��!��;�0D�GB�H,�Hj@��m"��3� C$R|"� `�T�O$7��!?>��@0DBɇ!(�H4�0f@� ,l"��3� C��|&� `�$��7��!�~>��0Db�G�H>�`�@��N�t"��ØC$W}&� `��Ϥ:��X��!�>�0D�ه�H`�0b��mFl�"��Ð C$�}� `��6�x��!��>��0D�gb�H3�Ln@�Ɍ�n"�� C$T~,� `��#6P��a� �!�=?��0D��H.�Lv@�)�� o"��C� C$�~(� `��ۏ�7��@��!�?��0D��!�Hd�0�@� Ο u"��3�C$[(� `��5�@��!��?�|0Dj��$6��$2�(L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B��1��G�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align&0��v#��#��#� #�ڍ ��8� ��9� ��9� ��:� l7��0��0��0�0�0��Y��[��\��^��_��Bi��j��m��m��x��x��z��|��v#��#��#��#�A#�Q#��#��#�R#�R#�"R#�2R#�BR#�RR#�bR#��R#"�R#%�R#(�R#+�R#.S#1S#9CS#<SS#?cS#��S#��S#�V#�V#�"V#�2V#�BV#��V#"�V#%�V#(�V#+�V#.�V#1�V#9#W#<3W#?CW#��W#��W#��W#��W#�Z#�Z#�"Z#�rZ#"�Z#%�Z#(�Z#+�Z#.�Z#1�Z#9[#<[#?#[#�[#��[#��[#��[#��[#��[#�^#�R^#"c^#%s^#(�^#+�^#.�^#1�^#9�^#<�^#?_#�_#�_#�_#�_#��_#��_#��_#�2b#"Cb#%Sb#(cb#+sb#.�b#1�b#9�b#<�b#?�b#�bc#�rc#��c#�c#��c#��c#��c#�f#"#f#%3f#(Cf#+Sf#.cf#1sf#9�f#<�f#?�f#iCg#�ۍ �M�� p7��6v6��6w6��B�� w#u�g#w�g#yj#{j#}#j#3j#�Cj#�Sj#��j#��j#��j#i#k#k3k#mCk#oSk#qck#ssk#u�k#w�k#y�k#{�k#}n#n#�#n#�3n#�cn#�sn#��n#io#ko#m#o#o3o#qCo#sSo#uco#w�o#y�o#{�o#}�o#�o#�r#�r#�Cr#�Sr#�cr#i�r#k�r#ms#os#q#s#s3s#uCs#w�s#y�s#{�s#}�s#�s#��s#��s#�#v#�3v#�Cv3�0�@�C3Fp�0 �1Ð��3K��043��03P�0HA4�0�D3U�0XA4�p�D3Y�0hA4ð�D3]�0xA4��`0��0�A3c��`0�P�0�A3g��C��C��C��C��C��C��(�C(��(��(�)�C)��)��)�*�C*��*��*�+�C+��+��+�,�C,��,��,�-�C-��-��-�.�C.��.��.�/�C/��/��/�8�C8��8��8�9�C9��9��9�:�C:��:��:�;�C;��;��;�<�C<��<��<�=�C=��=��=�>�C>��>��>�?�C?��?��?�H�CH��H��H�I�CI��I��I�J�CJ��J��J�K�CK��K/1�AL�0�D33��CM41�`�M�0�Dp39��N11ðAL�0�D@3=��O71��M�0�E@3a��X71�0�K�0�E3e��Y41�pM�0�Ep3i��Z/1ðAL�0�E3m��[41��M�0�Ep3q��\71�0�M�0�E�3u��]11�pM�0�E@3y��^71ð�K�0�E3}��_41��M�0�Fp3��h41�0�M�0�Fp3��i11�pAL�0�F@3��j71ð�M�0�F�3��k11��M�0�F@3��l71�0M�0�Fp3�� b�h�b�h�h h�h�b�h�h�h�h�h�h�h h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�b�b�h h�b�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h��&(7;��49�2��7�Q��H��ή�%��.�.��mnհ��Ѝ�� 0��@��P��`��p��Џ��!�1�A�Q�a�q��ё��0�1!�21�3A�4Q�5a�6q�7��8��9��:��;��<ѓ=�>�?�P�Q!�R1�SA�TQ�Ua�Vq�W��X��Y��Z��[��\ѕ]�^�_�p�q!�r1�sA�tQ�ua�vq�w��x��y��z��{��|ї}�~��!��1��A��Q��a��q��љ�ᙞ�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d�}\�}��4Ma A,$�� d�}\�}��4Ma A,T�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,D�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,t�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� d�}\�}��4Ma A,d�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,t�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,4�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� 6C�A8 �}\�}`�}��4M~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,$�� 6C�A8 �}\�}`�}��4M~��4Ma A,T�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,T�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma Ĉ�� G� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` ��kBp�p r2b` �nBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a Ĉ�� ć� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �kBp�p r2b` ��kBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a �� ć� �� ǬB�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ�ұ�� T7n��4խ��!8Mu�~0�`��4��DF~@̀��A!8Me[�A!8M��A!Q}��A!8M�[�!8Me��4U]��!8MU�`��ܷ��Txo��4U��AX�A��A ��!�n��,n��!˂�`NS巡@3Դ��T�m��4��!�a �� ć/ �� /B�@�}�}�}�}`��,>Ra �� G/ A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��Nd��4��!�r�F~@̀��A!8Me��A!8M�[�A!8M�[�!8Me��4U]��!8MU��~`NSպ��T5o��tD�m�`�A`��`��tD��,�m��5m��4�}��!�a �� ǫ �� B�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ��,�~`NSٶ�ұ�� T7n��4խ��!8Mu�~0�`��4��D6~PNS�v`NSὕ��T��@3�~PHGT�~@NS��A!8M�[��!8MU׶~`NSո��T�m��4U��AX�A��A ��!�n��,�}��!˂�f`��༡@3Դ��T6a �� < �� G<B�@�}�}�}�}`��,>Ra �� ; A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��T�n��,˭�4n��4�m��4Un��4o��A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o�`�A`��`��tD��,�m��5m��4�}��!�a Ĉ�� ć� �� Ǭ�pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a Ĉ�� ǫ �� pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�R d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��,��[�A!8M��A!8Mś�A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�S d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� ć� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��4Un��4o��tD�m��,��[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!Q}��A!8M��!8Me��4U]��!8MU�~PNS��`NS�q�u�y��`��"��Tym��4U��"��Tyo�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� Ǹ �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!8M��A!8M��A!Q}��!8Me��4U]��!8MU��`NS��T8o�`�A`��`��4U��!�n��,�� !8M��~0�`�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a �� ć� �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� ć/ �p1��Q�D8~0�c�`�A`��T��T�}`��a �� ǫ �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� < �p1��Q�D8~0�c�`�A`��4�n��T�}`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��kBp�p r2b` �!nBp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��kBp�p q1b` �!nBp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �kBp�p r2b` ��kBp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �kBp�p q1b` ��kBp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a ��A� ��}\�}@��a ��A� 쇙��}\�}@��a ��A� ��}\�}@��a ��A� 䇘��}\�}@��a ��A� ��}\�}@��a ��A� 쇏��}\�}@��a ��A� ��}\�}@��a ��A� 쇎��}\�}@��a ��A� ��}\�}@��a ��A� 䇍��}\�}@��a ��A� ��}\�}@��a ��A� 쇌��}\�}@��a ��A� ��}�}@��a ��A� ��} �u�,a ��A� �G��} �u�,a ��A� �Ȋ��} �u�,a ��A� �G��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a �� Ȉ A�@�}�}�}@��a �� H� A�@�}�}�}@��a �� A�@�}�}�}@��a ��A� ��} �u�,a ��A� ��~��} �u�,a �� H~ A�@�}�}�}@��a �� } A�@�}�}�}@��a ��A� �}��} �u�,a �� | A�@�}�}�}@��a �� H| A�@�}�}�}@��a ��A� ��{��} �u�,a ��A� �{��} �u�,a ��A� ��z��} �u�,a ��A� �z��} �u�,a ��A� ��y��} �u�,a ��A� �y��} �u�,a ��A� ��x��} �u�,a �� Hx A�@�}�}�}@��a ��A� ��o��} �u�,a ��A� �o��} �u�,a ��A� ��n��}�}@��a ��A� �n��}�}@��a �� m A�@�}�}�}@��a �� Hm A�@�}�}�}@��a ��A� ��l��}�}@��a ��A� �l��}�}@��a ��A� ��k��}�}@��a ��A� �k��}�}@��a ��A� ��j��}�}@��a ��A� �j��}�}@��a ��A� ��i��}�}@��a �� Hi A�@�}�}�}@��a ��A� ��h��}�}@��a ��A� �h��}�}@��a ��A� ��_��}�}@��a ��A� �_��}�}@��a ��A� ��^��}�}@��a �� H^ A�@�}�}�}@��a �� ] A�@�}�}�}@��a �� H] A�@�}�}�}@��a �� ] B�A8�}�}�}0@�c��,a ��A� �\��}�}@��a ��A� Ȉ[��}�}@��a ��A� �[��}�}@��a ��A� ��Z��}�}@��a ��A� �Z��}�}@��a ��A� ȈY��}�}@��a ��A� �Y��}�}@��a ��A� ��X��}�}@��a �� HX A�@�}�}�}@��a �� X B�A8�}�}�}�}@��a �� HO A�@�}�}�}@��a �� N A�@�}�}�}@��a ��A� �N��}�}@��a ��A� ��M��}�}@��a ��A� �M��}�}@��a ��A� ��L��}�}@��a ��A� �L��}�}@��a �� L B��@�}�}@H��!<Sa �� ؈K B��@�}�}@H��!<Sa ��A� ��J��}�}@��a ��A� �J��}�}@��a �� I A�@�}�}4�}@��a ��A� �I��}�}@��a ��A� ��H��}�}@��a �� HH A�@�}�}4�}@��a ��A� ��?��}�}@��a ��A� �?��}�}@��a ��A� ��>��}�}@��a ��A� �>��}�}@��a ��A� ��=��}�}@��a ��A� �=��}�}@��a ��A� ��<��}�}@��a �� < A�@�}�}�}@NS��!�a � �� <B�@�}�}�}@NS��!�a �� ; A�@�}�}�}@��a �� ; A�@�}�}�}@��a "D,��A|Ě\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� : A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� N A�@�}�}�}@��a �� 8 A�@�}�}�}@��a � �� H8B�@�}�}�}@NS��!�a �� H8 A�@�}�}�}@NS��!�a �� G/ A�@�}�}�}@��a �� G/ A�@�}�}�}@��a "D,��A|К\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� I. A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� HO A�@�}�}�}@��a �� + A�@�}�}�}@��a � �� +B�@�}�}�}@NS��!�a �� + A�@�}�}�}@NS��!�a �� * A�@�}0T�b�0!I�}@��a �� * A�@�}0T�b�0!I�}@��a ��A#�` &�4��}0T�b�0!I�}04�a��,a ��A#�` &�4��}0T�b�0!I�}04�a��,a �� ) A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� ,g4A�@�}0T�b�0!I�}p�a��,a �� ,e4A�@�}0T�b�0!I�}p�a��,a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� B�A8�}0T�b�0!I�}p�a��,a �ȈA� � � ��@�@ �}0�c��4�u��T~PNS�~PNS��}@��a �ȈA� �� @�@ �}0�c��4�u��T~PNS�~PNS��}@��a ;J,@ �� a��pC�`0�0�� %�1K�7�2F0bp p�C��#�`'?�,�,�1PAD!��@y(�p�$��A8V~`��Խ��A!0OM��!8̀Ӷ~`N3൝�,7m��,wo��<vm��4�q�u�y��,a 9J,@ �� ܐQ�$� E��,�@#�`�#8��p�Q��,Ca#�`�#9�$�ǈ�� H-�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~`N3ؼ��6~@̳��!8�`�f~`��ص��AX�A��!�b�V~P�S�&~@��a 9J,@ �� ȐQ�$� E��,�@#�`�#3��p�Q��,Ca#�`�#4�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛��!8�`�&~@��a 9J,@ �� Q�$� E��,�@#�`�#.��p�Q��,Ca#�`�#/�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛�!�b��4��a 5J,@� ��,C #�`�&D�$� ��,�@#�`�'d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�!D�$� ��,�@#�`�"d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |R?�,�1bP |R?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |?�,�1bP |?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a � �X�@�}�}@0�@��!8Ma � �T�@�}�}@0�@��!8Ma � �`�@�}�}@0�@��!8Ma �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �ȈA� �I��}@NS��D�`��,a �ȈA� ��}@NS��D�`��,a �ȈA� �I�� @�@ �}0�c��4��T��T�qOD�}@��a �ȈA� �� @�@ �}0�c��4��T��T�qOD�}@��a ��A� � ��}0<Sa ��A� ��}0<Sa ��A� � ��}0<Sa ��A� ��}0<Sa ��d�@�}<a ��d�@�}<a ��d�@�}<a ��d�@�}<a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a �� ̇�B�A8�}`8E!�q )�}@�Da ��A� ��} a �� J��}0�d�2a ��A� ��}@0��a �pCPa0bp ��6�� Ap�A8�}@<1��a A,@ � �� B0��}@<1��a �pCPa0bp ��6C� Ap�A8�}@<1��a A,@� �� B0��}@<1��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d �}PX�0F~ 8��ұX��!EB 4�q�� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| d�}PX�0&~PHT�#�}pH�P a�CDF~@8��A!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ �}PX�0~PHT�#�}0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u�L��a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��@,`��ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u �& )&~0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��4� H��N�#6~PH��#a &A,@�A5��<��9��9��A��6!P�\Ď�� d �}PX�0F~ 8��ұX��!EB 4�q�� Qa #A,@Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa %A,@�A5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u�L��a &A,@�A5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u �& )&~0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�}pH�P a��m��t,>b��HLDH�2Ֆ~@8��!Q��Ԇ~PH��#V~��L�}0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~PHT�#�~`H��#��(��D�D-Sm��D>R��A!��AHLDH�2��P�u�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��f~��L��ұ��m��Hm��t,>b� 1!Q�d�CD�}P�2a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a UA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��Hm��t,>b� 1!Q�d�CD�}�,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m�� %��~�HT�#>R�"��m� 1!Q�T��!Q��ұ��X�AHLDH�2��P��N�#>a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@d�}PX�0�~�HT�#>R��"��m� 1!Q�T��!Q��$��D�D-�qY�A1��!��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��X�AHLDH�2��!��qY�A1�X��D>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��!��$��D�D-�qY�� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#��uR$�@C�A"��X��D>�#>b� 1!Q��#>b�CDv~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0�~�HT�#��$ұ��4��D�D-��qY�A1��A"Q��"��>�#V~��L��a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��@,`��D>�#>b��#>b��t,>b��HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��4� H��$��$N�#��(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұ��!EB 4�u�� Q�q )a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#�}pH�P a��ұ��u��a A,@�&!P�܈A� �J� �a� �A8 �}PX�06~0H�b�CD�}P�2�}@0�� Qa A,@�&!P�܈A� �J� �a� �A8�}PX�0&~PHT�#6~PH��#�}0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұX��P�y�L��a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#F~@8��A!��X��P�y�L�q )a A,@�O6!P\Ď$��B$H0��}PX�0V~ 8��ұX��P�q )�}�,@C�� Qa A,@�O6!P\Ď$��B$H0�� }PX�06~PHT�#V~@8��A!��X��P�q )�}�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d�}PX�0F~ 8��ұ��!EB 4��,�m�CD&~0HT`�H��,a A,@�&!P�܈A� ��# �`p+.��!�&��A� �� }PX�0v~`��Է�ұX��P�y�L�q )&~0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d�}PX�0F~ 8��ұ��!�R��}0T@d��@,`�H��X�!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8�}PX�0V~ 8��ұX��!�R��}0T@d�H�y �& )6~0HT`��,a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�yR$�@CX��D>b� 1!Q�d��D>R[��Pa MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a FA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��t,>b�H��Hm� 1!Q�d�CD�}P�2a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��,��D�D-��!�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��m��t,>b�H��Hm��$��D�D-�u��A1�a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a JA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�y �& )�~`HT�#��$��D�D-��N�#�}0T@a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a MA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�yR$�@C�"Q��Hm��t,>�#��$��D�D-�u��!��a GA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ~�E�@�}PX�0f~��L��b��D>�#�� ұ��V~��L�}0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�� ~�H��#>b[�AHLDH�2Y��P�y�L��ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�vsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@�}PX�0v~��L��b��4� H��N�#>b��~�HT�#>R��"��m� 1!Q�d�CD�~pH��#>a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�~�HT�#��yR$�@C�A"��X�AHLDH�2��X��D>�#>b�CD�}@0��A"Q��A�D>r��X��"��>�#f~��L��a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��ұ��X��HLDH�2�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0�~�HT�#��$ұ��4��D�D-��u��A1��b��D>�#>b��t,>b��HLDH�2�>�#a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d��@,`�H��$��$N�#��(ұ��#>b� 1!Q�T��a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a PA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d�H�y �& )�~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8MU�&~@NS�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�6~��L��T5n��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pn0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pnPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��pn`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��pn`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,�}7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_�_�H#�d��4�m��A��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�F~@NS��HLDH�2�V~`NSո��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,�}7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!�~��4M~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �q #2"�\�.�E�^�.��`�.��b�/86��d�/XF�f�0x� �h�0�� j�0�6�l�1�F�n�1��p�1�)�r�287-�t�2XG8�v�3x�:�x�3��=�z�3�7I�|�4�GL�~�4�7O��5@Z��58�\��5X�_��6x0k��6�@n��6�0y��7��|��8��98`��:XP��:x௃��;��;��ȃ��<�0؃��=�0郠�?��?8��@X��Ax��B��.��C��?��D��X��E��[��E�k��G8 ��Hh ��H��I� ��I(��Ih ��I��I� ��J��JH ��J��J�0��JЯ��KHP��K��K��K��KH��L�Ʉ��L��ʄ��L( ̄��LH(�̈́��L�(0τ��M�(�؄��M)@ڄ��MH)�ۄ��M�)@݄��M�)�ބ��M*P脢�NH*�鄦�N�*p넪�N�*0턮�N+��NH+0��O�+��O�+0��O,��OH,��O�,��P�,� ��P- ��PH-��P�-P��P�-��Q.`��QH.��Q�.`��Q�.��Q/��R(/`)��Rh/�-��S�/`:��S�/<��S�/�=��S�/@?��T(8 I��TH8�J��T�8@O��U�8�[��U�8`]��U�8_��U9�h��V89�j��Vx9`l��V�9�n��V�9`x��W:@z��WX: |��W�:~��W�:��X;��XX;��X�;��X�;��Y<��YX<`��Z�<ૅ��Z�=@��[8> ʅ��\�>م��]x? ��^�?��^�?P��^H0텂�^HH�_�H��_�H��_I��_HIp��_�I0 ��`�I� ��`J��`(J ��`HJ0��`hJp��a�J��a�J@��aK��a8K��aXK(��bxK�+��b�K�8��c�KP=��c�K`J��d�KpO��eL0\��e8L l��gXL�~��hxL`��i�L`��k�L��l�LPچ��m�L��oM��p8Mp��rXM�.��s�M`>��t�M�I��t�M@M��t�M�X��u(N]��uHNpi��vhN�x��w�N��x�NP��z�N��{�N྇��|OP؇��}(O0��~HO��hO� ��O��͂�O�.��O�9��O`=��X�H��(X�X��HX�_��hXo��X�{��؇�X ��X��XЛ��Y��(Y0��HY��hY��Yp��Y��YP��Y��Z��(Z� ��ݐHZ��БhZ��Z*��Z�.��Z�>��Z�M��[ ]��([�h��H[�o��h[`|��[��[@��Θ�[��[��\��(\��H\��h\��\��\��ޞ�\��џ�\��]p��(]��ڡH]P��h]�*��Ȣ�]0.��]�9��ˣ�]�I��Ȥ�]�X��^�_��(^�l��H^�o��h^�|��^��^Л��^��˪�^`��_��(_p��ŬH_P��h_0��_��_`��_��_� ��h��α(h��Hh,��hhP<��hL��h�[��ٵ�hp_��h o��i@|��(i ��Hi��ݸhi��i0��i��i��i��j��(j��Hj ��hj��j� ��j@��j�(��j0.��k@:��(kP>��HkPJ��hk�Y��k�h��k0o��k |��k ��l@��(l��Hl0��hl��l��l��lp��l��m��(m��Hm��hm��mp��m`��m`+��m�8��n�?��(npN��Hn ]��hn i��n�o��n�|��n��nЌ��oИ��(oP��Ho��ho��o��o��o��o��x��(x`��Hx`��hx��x��x,��x8��x<��y�K��(y@Z��Hy�h��hy�m��y�y��y��y@��y��z��(z��HzP��hz@��z`�]��threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_50_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��U( ��EDT��*�Re�H)#�E!�-��B��S�0�� A9@S��L@W��~ʪ�@[��@�WE0� �T@c�E��,2��(BhQZG�X��"?��[�)RE�"�Ȓ"XB�E�!��/��E{E@�/��(�Q`H��F�ЈateR�L#ň�"��5F��"��IF��ʨ2��\F�Q2EHF3ڌ�):��ge��2\Q$Z�(�"�W(�V�+�*�P+�k�A�� pE\��ke��.K�2\qW($[��K��W��V�+��p+��B�� !��pEb��oe��1R�2\�X5$\�H�"�WPFW�+.��+��%Cʕ��!��`�(b�*�E��]W�ȡ�@Z��!��Z��E��;��E��:�ȼ"tw�g(�zz�zE��/j��F��(�"�g�W��ct_�{��3�/�a�ˈ?C��/��?��?#�CN�,�@� ��,�k��;� �qE\w�q��`!��b\qAX�+�2��W�Da1��,,��qaX�+��;� qEb@� ��B\��_~!�q�f��/Db!��r�D,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8no�P�� @��av@�1�7��P�av@h��Aw�B"�� F.�� @��\0��`�2�`y7��|r�8�n ��q��@;��ແv@��w�R-�� ].�� @�ڿ\0��`D�A.G� @�j� �C� ��h��1z�~5�� v@x�q��@;�x�8�n �P;��7��Y@��`��- �� @��@�� }� �qz�FPp�@��v@�(��q��@;��\�8To �P{ �`�7�hOW0��`��+G� @��]�� z��Yp〽�v@�-��q��@;�\�8do �P��`�7�h�W0��`��+G� @�:�� "��-@��r� �H�9xC�� `��%0D��!��"��Y@�`u�, �H�:`C$&��!R��@� n�� `�t�C@��w��!��0D��2 �Hq<hf@�i� �"��0h@��K �"��C��8 `��D@�v{��!|�0DR�� Ha>px@��χ+�"��C$x�Q `�T�6 @�~~�J��!��\)0D��: �H�?p�@�� `��C�Y$�`�"�$��C��$�W `��@�ȓ�f��!R�V,0D�Q� �HoJ�@��N �"�)��C$Y%�^``�d�� @�J�x��!��8P0D"[2�HtKD��t��@"�.!� C$�%$y@`��T,�Z��!R?P0Drc"��HLH"��)� �$@"�31� C${&��@`��,�z�hb��!�ZM0D�k¹ �H�M@:@�鶉�'"�7�C��& � `�t�T�r�x��!��Z0D�sr�H�NDq@�)׉�."�;!�C$~'$� `��B��!R�i0D}B: �H�OL�@��5"�?A�C$,�� `�$�Ee�r��!+o0D�� H�X\�@��%��<"�e!�C$�,&� `�T��r��!RzC|0Db�B��H�Y@�@��E�!>�"�h!�C�1-�� `��3��`��!R�H|0D��H�Z0�@�i]g>"�k��C��-�� `�Բ�3�v��!��|0D��"��H�[L�@�i|g>"�o�C�.�� `��4�~�p��!R�|0D��H�\@�@�i��i>"�sQ�C��.�� `�d�E5�ֺ`��!�]L}0D킩�H�]0�@�i��>"�w��C��.�� `�T�s�f�X��!��J~0DB�b��Hs^$�@��ϋe?"�z��C�X/�� `��E�@�̽`��!��~0D��H&_$�@��$?"�|��C$�/�� `��ű@��H��!R��~0D �"��H�_ �@�I��d?"�~q�C$�/�� `��@��8��!��E0D��H&h�@��?�"ɠq�C�4� `�ԃ��@��!R�0D�C#��H{h�@�) �?"��Q�C$W4� `�d�1"@�t�@J��!�8J�0DbG9�H�h$'@��?"��A�C$�4�� `��@�n�HF��!Xň0D*KCQ�Hoi*@��.��?"��A�C��4 `��"@�`�PJ��!�nF�0DO��H�i�@��=��?"ѧA�C$�4�� `��F�@�J�@��!R��0D�Q��H9j�@��G �?"��Q�C$55�� `�$�F�@��PF��!R�ƈ0D:U��H�j�@�IV �?"�Q�C�_5�� `�t�F�@��H��!R�F~0D�Wc)�H&k%@��e �E"嬁�C��5�� `��F�@��H��!��0D�[��H�k�@�Iu��?"ͮa�C��5�� `�$��@��H��!R��0D�_#��H l�@�Ʉ eD"��a�C$6�� `��F�@��8��!��0Drc��Hvl�@�i� �?"��q�C$V6�� `�d��@�z�hF��!�:Ɉ0D�g��H!m(#@�i� gD"崡�C��6 `��#@��H��!h�0D2mC��H�m2@�鶍EF"��C$�6�� `�4߆�@��@��!R��0D2q#��H/n2@�� EF"��q�C$37�� `�4��@��HF��!��ň0DZu#��H�n4@� ׍dD"��Q�C$s7� `��F1"@��H��!R��0Dy#��H,o�@�I��?"��a�C��7�� `��@��H��!R��0D�{��H�o$6@� � �F"��q�C��7�� `�4��@��0��!��0D��H&x8@�)�?"��1�C$<�� `��@�v�8p��!��0D*�C��Hwx8�@��?"��C�r<�� `��@��X��!RH�0DJ��H8y,�@�i,��?"��C$�<�� `��@�J�ht��!�l�0D�ͣ��H�y�@�i=$>"��C��<0 `�4�G&@�P�@��!��0D��H>z `@��LL"�遀 C$;=0 `��&�L�@F��!��0D�Ճ� �H�z >@�I\DG"��A�C$w=�� `�d��@��!��0DZ��H3{`@�)gL"�큀 C��=0 `��&��@��!R��0DJ��H�{ <@�)w�G�"��C��= `�Ծ�#&��`��! ��0D�� Hc|<b@�I��GL"�� C$Y>1 `��G$&��h��!R?�0Dr�! �H;}<b@�I�FL�"�� C��>1 `��#&��`��!{��0D�� H'~<b@��FL�"�� C�8?1 `��#&�X�`��!��0D:�� H�~8b@�)�EL�"�� C$�?1 `�T��#&��`��!��0Dz�� H�@C��?��!��@��f `�4��c&��q��!"��0D E�1 �H��@h@� M�"�#r� C�zD 6 `��ȡ&0�B)��!�I"��0D�I� �Hf� p@��-N�"�%�� C$�D09 `�d�H�#@�~Ix��!z" �0D�O$� �H��`r@�IF��G�"�(��C$2E `�ԦH�#@��!��" �0D�U$� �H��$<@�)]��G�"�+�� C$�E�� `��ȳ@��y��!R�"�~0Db]��H�@�@�ih?�"�0�C�F�� `��g'�Nq��!R-#�0Dg�� H�8x@��N"�4�� C$�F< `��ȃ'@�Ty��!Rn#��0D�o�� H#�8|@�)��O�"�8�� C�6G<= `��'@�ty��!R�#�0D�w�� H'�x~@��O"�>�@��L"�? C��G3 `�䂉c&@�x0y��!�&��0D��1�H��4f@��L"�c C�wL3 `��Ic&@�V2y��!�N&��0DB��1�Hw�8f@��4��L"�f� C$�L3 `��Ic&@��3y��!�&��0D��D�Hd� l@��MKT "�j2� C�ZM(5�`��I�&@��5��!R�&�0D��$��Hn�d�@��t�("�n�� C��M�8 `�D� F*@4�x8y��!�'��0D��3�H�|�@�)��00�Hb�x�@� ��"�v��C$�N�Z `�� W��<Y��!�'�|�0Dj�dB�H��X��MU"�|�� C$�OT `��Ʌ*@<�`?��!�'S�0D��H#�@�@�i�)>"��I��Pa��!�(*�|0DzE��H��<�@��i>"��"�C��T�� `�$�J4��Ra^��!�_*ϫ0DrMEz�H�D�@�)>&V"�� C�U�X `��+��T�b��!R�*R�0DZU%��H��D�@��]fV"�� C$�U�Y `��J4+��Vab��!��*L�0D]��H��<�@�)|)V"ͯ"� C�V�X `�d�J+��X�j��!�+׭0D�ee��H�X�@�I��V"�� C$�V�Z `��U+�n[�j��!�{+S�0D q�� H/�8�@��+W"͹�� C�UW�\ `�$��+��]�z��!�+V�0D�y��Ht�`�@��W"��r� C��W�^ `�$��+��p�~��!�.د0D��H�L�@��W"��b� C$�\�^ `�ė�+�hs�z��!{.W�0D��H:�,�@��L�%\"�鲌C�=]�q `�d��b.D�buY �H��,�@��\(]"��C��]�w `��K$/��vy��!�.�0D:݅��H��,��|��\�"��@��wY��!� /��0D��% �Hi�H$@�i��d"��C�^^�� `��΋3�FzYd��!�K/K�0D��%��Hj�D;@�i��g"��C��^�p `�T�3.D��{a��!R/��0DJ�"��C�0_�t `�$�K�.�F}��!R�/��0D:��H��@�@��j_"��C$�_s@`�D�0D�� H��H�@��)d"��2�C�d�� `�d�L�2�ΐ�Z��!2�0DzE��H��05@�� g"�##�C��d&� `��.�Ԓa��!�]2̸0DM�"�H��0�@�I7 `�d�.�䓉��!҈2�0D�QFz�Hf�H�@� O�^"�*�C$\e�} `�Ԯ3.H��a� �!R�2@�)f&\"�,#�C��e�� `��D2�h��N��!��2��0D�_&j�H)�\0@� �Ff"�1��C$<f$� `��ʌ�3�x��|��!�;3L�0D�g� �H'�0�@�i�f\�"�4ØC��f��!�]3L�0Dm&J�H��D�@��]"�7#�C�g�z `��/�Ҝ��!��3̸ 0DRu�1$�H��0C$qg�p `��/�@��B��!��3�0D*{f:�Hx�D*@�)��e"�>s�C��g� `��S3�R��p��!�6��0Dj�f��H��0�@��&\"�bÌC$rl�q `�� c.D��a �H(�0�@��&�(]"�e�C$�l�w `��$/�t��!�z6�0D�Ϧ��H/�0��iG�\�"�i�@��a��!ҟ6��0D��F �H��L$@��^��d"�l�C$�m�� `��3�D�ad��!��6L�0D��F��H��H;@��g"�pÄC�n�p `�� 3.D��a��!R(7��0Dj冁"�rÄC�rn�t `�d�M�.�N��!RN7��0DZ��H{�@�@�i��j_"�wÌC$�ns@`�� 0D"� �H0�H�@�I�)d"�y3�C�So�� `��M�2�ֽ�Z��!�7�0D��H?�05@�i� g"�}#�C��o&� `�$� 6�ܿa��!��7��00D2A�&�H2�0d��l"��À C$;t$� `��6�tљ��!R;:��0DI'��H1�Dr@�I,�n"�È C$�t��`��c6�x�a��!z:��0D�Og��H.�Pz@�IL�o"٩#� C$Tu$� `��&:��aH��!��:��0DRYgB�H=�L�@� n�u"��À C$�u� `�Ļ#6L��a�� !�:�00Da�1�H,�0`@� � m"��#� C$>v&� `��6��ى��!�<;��0D"iǢ�H:�0b�ɬ�l�"ɵØ C$�v� `��d7�zۙ��!|;��0D"q��H8�H~@�I� t"񹃉C$Zw� `��c:��ݙP��!ҿ;��0D�y�b�Hf�0`@�I�l"��È C$�w�``�d�C6L��a��!��;�0D�GB�H,�Hj@��m"��3� C$R|"� `�T�O$7��!?>��@0DBɇ!(�H4�0f@� ,l"��3� C��|&� `�$��7��!�~>��0Db�G�H>�`�@��N�t"��ØC$W}&� `��Ϥ:��X��!�>�0D�ه�H`�0b��mFl�"��Ð C$�}� `��6�x��!��>��0D�gb�H3�Ln@�Ɍ�n"�� C$T~,� `��#6P��a� �!�=?��0D��H.�Lv@�)�� o"��C� C$�~(� `��ۏ�7��@��!�?��0D��!�Hd�0�@� Ο u"��3�C$[(� `��5�@��!��?�|0Dj��$6��$2�(L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B��1��G�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align&0��v#��#��#� #�ڍ ��8� ��9� ��9� ��:� l7��0��0��0�0�0��Y��[��\��^��_��Bi��j��m��m��x��x��z��|��v#��#��#��#�A#�Q#��#��#�R#�R#�"R#�2R#�BR#�RR#�bR#��R#"�R#%�R#(�R#+�R#.S#1S#9CS#<SS#?cS#��S#��S#�V#�V#�"V#�2V#�BV#��V#"�V#%�V#(�V#+�V#.�V#1�V#9#W#<3W#?CW#��W#��W#��W#��W#�Z#�Z#�"Z#�rZ#"�Z#%�Z#(�Z#+�Z#.�Z#1�Z#9[#<[#?#[#�[#��[#��[#��[#��[#��[#�^#�R^#"c^#%s^#(�^#+�^#.�^#1�^#9�^#<�^#?_#�_#�_#�_#�_#��_#��_#��_#�2b#"Cb#%Sb#(cb#+sb#.�b#1�b#9�b#<�b#?�b#�bc#�rc#��c#�c#��c#��c#��c#�f#"#f#%3f#(Cf#+Sf#.cf#1sf#9�f#<�f#?�f#iCg#�ۍ �M�� p7��6v6��6w6��B�� w#u�g#w�g#yj#{j#}#j#3j#�Cj#�Sj#��j#��j#��j#i#k#k3k#mCk#oSk#qck#ssk#u�k#w�k#y�k#{�k#}n#n#�#n#�3n#�cn#�sn#��n#io#ko#m#o#o3o#qCo#sSo#uco#w�o#y�o#{�o#}�o#�o#�r#�r#�Cr#�Sr#�cr#i�r#k�r#ms#os#q#s#s3s#uCs#w�s#y�s#{�s#}�s#�s#��s#��s#�#v#�3v#�Cv3�0�@�C3Fp�0 �1Ð��3K��043��03P�0HA4�0�D3U�0XA4�p�D3Y�0hA4ð�D3]�0xA4��`0��0�A3c��`0�P�0�A3g��C��C��C��C��C��C��(�C(��(��(�)�C)��)��)�*�C*��*��*�+�C+��+��+�,�C,��,��,�-�C-��-��-�.�C.��.��.�/�C/��/��/�8�C8��8��8�9�C9��9��9�:�C:��:��:�;�C;��;��;�<�C<��<��<�=�C=��=��=�>�C>��>��>�?�C?��?��?�H�CH��H��H�I�CI��I��I�J�CJ��J��J�K�CK��K/1�AL�0�D33��CM41�`�M�0�Dp39��N11ðAL�0�D@3=��O71��M�0�E@3a��X71�0�K�0�E3e��Y41�pM�0�Ep3i��Z/1ðAL�0�E3m��[41��M�0�Ep3q��\71�0�M�0�E�3u��]11�pM�0�E@3y��^71ð�K�0�E3}��_41��M�0�Fp3��h41�0�M�0�Fp3��i11�pAL�0�F@3��j71ð�M�0�F�3��k11��M�0�F@3��l71�0M�0�Fp3�� b�h�b�h�h h�h�b�h�h�h�h�h�h�h h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�b�b�h h�b�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h��&(7;��49�2��7�Q��H��ή�%��.�.��mnհ��Ѝ�� 0��@��P��`��p��Џ��!�1�A�Q�a�q��ё��0�1!�21�3A�4Q�5a�6q�7��8��9��:��;��<ѓ=�>�?�P�Q!�R1�SA�TQ�Ua�Vq�W��X��Y��Z��[��\ѕ]�^�_�p�q!�r1�sA�tQ�ua�vq�w��x��y��z��{��|ї}�~��!��1��A��Q��a��q��љ�ᙞ�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d�}\�}��4Ma A,$�� d�}\�}��4Ma A,T�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,D�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,t�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� d�}\�}��4Ma A,d�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,t�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,4�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� 6C�A8 �}\�}`�}��4M~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,$�� 6C�A8 �}\�}`�}��4M~��4Ma A,T�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,T�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma Ĉ�� G� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` ��kBp�p r2b` �nBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a Ĉ�� ć� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �kBp�p r2b` ��kBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a �� ć� �� ǬB�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ�ұ�� T7n��4խ��!8Mu�~0�`��4��DF~@̀��A!8Me[�A!8M��A!Q}��A!8M�[�!8Me��4U]��!8MU�`��ܷ��Txo��4U��AX�A��A ��!�n��,n��!˂�`NS巡@3Դ��T�m��4��!�a �� ć/ �� /B�@�}�}�}�}`��,>Ra �� G/ A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��Nd��4��!�r�F~@̀��A!8Me��A!8M�[�A!8M�[�!8Me��4U]��!8MU��~`NSպ��T5o��tD�m�`�A`��`��tD��,�m��5m��4�}��!�a �� ǫ �� B�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ��,�~`NSٶ�ұ�� T7n��4խ��!8Mu�~0�`��4��D6~PNS�v`NSὕ��T��@3�~PHGT�~@NS��A!8M�[��!8MU׶~`NSո��T�m��4U��AX�A��A ��!�n��,�}��!˂�f`��༡@3Դ��T6a �� < �� G<B�@�}�}�}�}`��,>Ra �� ; A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��T�n��,˭�4n��4�m��4Un��4o��A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o�`�A`��`��tD��,�m��5m��4�}��!�a Ĉ�� ć� �� Ǭ�pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a Ĉ�� ǫ �� pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�R d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��,��[�A!8M��A!8Mś�A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�S d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� ć� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��4Un��4o��tD�m��,��[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!Q}��A!8M��!8Me��4U]��!8MU�~PNS��`NS�q�u�y��`��"��Tym��4U��"��Tyo�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� Ǹ �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!8M��A!8M��A!Q}��!8Me��4U]��!8MU��`NS��T8o�`�A`��`��4U��!�n��,�� !8M��~0�`�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a �� ć� �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� ć/ �p1��Q�D8~0�c�`�A`��T��T�}`��a �� ǫ �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� < �p1��Q�D8~0�c�`�A`��4�n��T�}`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��kBp�p r2b` �!nBp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��kBp�p q1b` �!nBp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �kBp�p r2b` ��kBp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �kBp�p q1b` ��kBp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a ��A� ��}\�}@��a ��A� 쇙��}\�}@��a ��A� ��}\�}@��a ��A� 䇘��}\�}@��a ��A� ��}\�}@��a ��A� 쇏��}\�}@��a ��A� ��}\�}@��a ��A� 쇎��}\�}@��a ��A� ��}\�}@��a ��A� 䇍��}\�}@��a ��A� ��}\�}@��a ��A� 쇌��}\�}@��a ��A� ��}�}@��a ��A� ��} �u�,a ��A� �G��} �u�,a ��A� �Ȋ��} �u�,a ��A� �G��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a �� Ȉ A�@�}�}�}@��a �� H� A�@�}�}�}@��a �� A�@�}�}�}@��a ��A� ��} �u�,a ��A� ��~��} �u�,a �� H~ A�@�}�}�}@��a �� } A�@�}�}�}@��a ��A� �}��} �u�,a �� | A�@�}�}�}@��a �� H| A�@�}�}�}@��a ��A� ��{��} �u�,a ��A� �{��} �u�,a ��A� ��z��} �u�,a ��A� �z��} �u�,a ��A� ��y��} �u�,a ��A� �y��} �u�,a ��A� ��x��} �u�,a �� Hx A�@�}�}�}@��a ��A� ��o��} �u�,a ��A� �o��} �u�,a ��A� ��n��}�}@��a ��A� �n��}�}@��a �� m A�@�}�}�}@��a �� Hm A�@�}�}�}@��a ��A� ��l��}�}@��a ��A� �l��}�}@��a ��A� ��k��}�}@��a ��A� �k��}�}@��a ��A� ��j��}�}@��a ��A� �j��}�}@��a ��A� ��i��}�}@��a �� Hi A�@�}�}�}@��a ��A� ��h��}�}@��a ��A� �h��}�}@��a ��A� ��_��}�}@��a ��A� �_��}�}@��a ��A� ��^��}�}@��a �� H^ A�@�}�}�}@��a �� ] A�@�}�}�}@��a �� H] A�@�}�}�}@��a �� ] B�A8�}�}�}0@�c��,a ��A� �\��}�}@��a ��A� Ȉ[��}�}@��a ��A� �[��}�}@��a ��A� ��Z��}�}@��a ��A� �Z��}�}@��a ��A� ȈY��}�}@��a ��A� �Y��}�}@��a ��A� ��X��}�}@��a �� HX A�@�}�}�}@��a �� X B�A8�}�}�}�}@��a �� HO A�@�}�}�}@��a �� N A�@�}�}�}@��a ��A� �N��}�}@��a ��A� ��M��}�}@��a ��A� �M��}�}@��a ��A� ��L��}�}@��a ��A� �L��}�}@��a �� L B��@�}�}@H��!<Sa �� ؈K B��@�}�}@H��!<Sa ��A� ��J��}�}@��a ��A� �J��}�}@��a �� I A�@�}�}4�}@��a ��A� �I��}�}@��a ��A� ��H��}�}@��a �� HH A�@�}�}4�}@��a ��A� ��?��}�}@��a ��A� �?��}�}@��a ��A� ��>��}�}@��a ��A� �>��}�}@��a ��A� ��=��}�}@��a ��A� �=��}�}@��a ��A� ��<��}�}@��a �� < A�@�}�}�}@NS��!�a � �� <B�@�}�}�}@NS��!�a �� ; A�@�}�}�}@��a �� ; A�@�}�}�}@��a "D,��A|Ě\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� : A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� N A�@�}�}�}@��a �� 8 A�@�}�}�}@��a � �� H8B�@�}�}�}@NS��!�a �� H8 A�@�}�}�}@NS��!�a �� G/ A�@�}�}�}@��a �� G/ A�@�}�}�}@��a "D,��A|К\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� I. A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� HO A�@�}�}�}@��a �� + A�@�}�}�}@��a � �� +B�@�}�}�}@NS��!�a �� + A�@�}�}�}@NS��!�a �� * A�@�}0T�b�0!I�}@��a �� * A�@�}0T�b�0!I�}@��a ��A#�` &�4��}0T�b�0!I�}04�a��,a ��A#�` &�4��}0T�b�0!I�}04�a��,a �� ) A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� ,g4A�@�}0T�b�0!I�}p�a��,a �� ,e4A�@�}0T�b�0!I�}p�a��,a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� B�A8�}0T�b�0!I�}p�a��,a �ȈA� � � ��@�@ �}0�c��4�u��T~PNS�~PNS��}@��a �ȈA� �� @�@ �}0�c��4�u��T~PNS�~PNS��}@��a ;J,@ �� a��pC�`0�0�� %�1K�7�2F0bp p�C��#�`'?�,�,�1PAD!��@y(�p�$��A8V~`��Խ��A!0OM��!8̀Ӷ~`N3൝�,7m��,wo��<vm��4�q�u�y��,a 9J,@ �� ܐQ�$� E��,�@#�`�#8��p�Q��,Ca#�`�#9�$�ǈ�� H-�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~`N3ؼ��6~@̳��!8�`�f~`��ص��AX�A��!�b�V~P�S�&~@��a 9J,@ �� ȐQ�$� E��,�@#�`�#3��p�Q��,Ca#�`�#4�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛��!8�`�&~@��a 9J,@ �� Q�$� E��,�@#�`�#.��p�Q��,Ca#�`�#/�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛�!�b��4��a 5J,@� ��,C #�`�&D�$� ��,�@#�`�'d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�!D�$� ��,�@#�`�"d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |R?�,�1bP |R?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |?�,�1bP |?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a � �X�@�}�}@0�@��!8Ma � �T�@�}�}@0�@��!8Ma � �`�@�}�}@0�@��!8Ma �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �ȈA� �I��}@NS��D�`��,a �ȈA� ��}@NS��D�`��,a �ȈA� �I�� @�@ �}0�c��4��T��T�qOD�}@��a �ȈA� �� @�@ �}0�c��4��T��T�qOD�}@��a ��A� � ��}0<Sa ��A� ��}0<Sa ��A� � ��}0<Sa ��A� ��}0<Sa ��d�@�}<a ��d�@�}<a ��d�@�}<a ��d�@�}<a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a �� ̇�B�A8�}`8E!�q )�}@�Da ��A� ��} a �� J��}0�d�2a ��A� ��}@0��a �pCPa0bp ��6�� Ap�A8�}@<1��a A,@ � �� B0��}@<1��a �pCPa0bp ��6C� Ap�A8�}@<1��a A,@� �� B0��}@<1��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d �}PX�0F~ 8��ұX��!EB 4�q�� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| d�}PX�0&~PHT�#�}pH�P a�CDF~@8��A!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ �}PX�0~PHT�#�}0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u�L��a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��@,`��ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u �& )&~0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��4� H��N�#6~PH��#a &A,@�A5��<��9��9��A��6!P�\Ď�� d �}PX�0F~ 8��ұX��!EB 4�q�� Qa #A,@Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa %A,@�A5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u�L��a &A,@�A5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u �& )&~0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�}pH�P a��m��t,>b��HLDH�2Ֆ~@8��!Q��Ԇ~PH��#V~��L�}0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~PHT�#�~`H��#��(��D�D-Sm��D>R��A!��AHLDH�2��P�u�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��f~��L��ұ��m��Hm��t,>b� 1!Q�d�CD�}P�2a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a UA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��Hm��t,>b� 1!Q�d�CD�}�,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m�� %��~�HT�#>R�"��m� 1!Q�T��!Q��ұ��X�AHLDH�2��P��N�#>a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@d�}PX�0�~�HT�#>R��"��m� 1!Q�T��!Q��$��D�D-�qY�A1��!��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��X�AHLDH�2��!��qY�A1�X��D>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��!��$��D�D-�qY�� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#��uR$�@C�A"��X��D>�#>b� 1!Q��#>b�CDv~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0�~�HT�#��$ұ��4��D�D-��qY�A1��A"Q��"��>�#V~��L��a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��@,`��D>�#>b��#>b��t,>b��HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��4� H��$��$N�#��(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұ��!EB 4�u�� Q�q )a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#�}pH�P a��ұ��u��a A,@�&!P�܈A� �J� �a� �A8 �}PX�06~0H�b�CD�}P�2�}@0�� Qa A,@�&!P�܈A� �J� �a� �A8�}PX�0&~PHT�#6~PH��#�}0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұX��P�y�L��a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#F~@8��A!��X��P�y�L�q )a A,@�O6!P\Ď$��B$H0��}PX�0V~ 8��ұX��P�q )�}�,@C�� Qa A,@�O6!P\Ď$��B$H0�� }PX�06~PHT�#V~@8��A!��X��P�q )�}�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d�}PX�0F~ 8��ұ��!EB 4��,�m�CD&~0HT`�H��,a A,@�&!P�܈A� ��# �`p+.��!�&��A� �� }PX�0v~`��Է�ұX��P�y�L�q )&~0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d�}PX�0F~ 8��ұ��!�R��}0T@d��@,`�H��X�!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8�}PX�0V~ 8��ұX��!�R��}0T@d�H�y �& )6~0HT`��,a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�yR$�@CX��D>b� 1!Q�d��D>R[��Pa MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a FA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��t,>b�H��Hm� 1!Q�d�CD�}P�2a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��,��D�D-��!�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��m��t,>b�H��Hm��$��D�D-�u��A1�a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a JA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�y �& )�~`HT�#��$��D�D-��N�#�}0T@a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a MA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�yR$�@C�"Q��Hm��t,>�#��$��D�D-�u��!��a GA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ~�E�@�}PX�0f~��L��b��D>�#�� ұ��V~��L�}0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�� ~�H��#>b[�AHLDH�2Y��P�y�L��ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�vsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@�}PX�0v~��L��b��4� H��N�#>b��~�HT�#>R��"��m� 1!Q�d�CD�~pH��#>a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�~�HT�#��yR$�@C�A"��X�AHLDH�2��X��D>�#>b�CD�}@0��A"Q��A�D>r��X��"��>�#f~��L��a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��ұ��X��HLDH�2�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0�~�HT�#��$ұ��4��D�D-��u��A1��b��D>�#>b��t,>b��HLDH�2�>�#a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d��@,`�H��$��$N�#��(ұ��#>b� 1!Q�T��a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a PA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d�H�y �& )�~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8MU�&~@NS�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�6~��L��T5n��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pn0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pnPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��pn`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��pn`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,�}7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_�_�H#�d��4�m��A��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�F~@NS��HLDH�2�V~`NSո��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,�}7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!�~��4M~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �q #2"�\�.�E�^�.��`�.��b�/86��d�/XF�f�0x� �h�0�� j�0�6�l�1�F�n�1��p�1�)�r�287-�t�2XG8�v�3x�:�x�3��=�z�3�7I�|�4�GL�~�4�7O��5@Z��58�\��5X�_��6x0k��6�@n��6�0y��7��|��8��98`��:XP��:x௃��;��;��ȃ��<�0؃��=�0郠�?��?8��@X��Ax��B��.��C��?��D��X��E��[��E�k��G8 ��Hh ��H��I� ��I(��Ih ��I��I� ��J��JH ��J��J�0��JЯ��KHP��K��K��K��KH��L�Ʉ��L��ʄ��L( ̄��LH(�̈́��L�(0τ��M�(�؄��M)@ڄ��MH)�ۄ��M�)@݄��M�)�ބ��M*P脢�NH*�鄦�N�*p넪�N�*0턮�N+��NH+0��O�+��O�+0��O,��OH,��O�,��P�,� ��P- ��PH-��P�-P��P�-��Q.`��QH.��Q�.`��Q�.��Q/��R(/`)��Rh/�-��S�/`:��S�/<��S�/�=��S�/@?��T(8 I��TH8�J��T�8@O��U�8�[��U�8`]��U�8_��U9�h��V89�j��Vx9`l��V�9�n��V�9`x��W:@z��WX: |��W�:~��W�:��X;��XX;��X�;��X�;��Y<��YX<`��Z�<ૅ��Z�=@��[8> ʅ��\�>م��]x? ��^�?��^�?P��^H0텂�^HH�_�H��_�H��_I��_HIp��_�I0 ��`�I� ��`J��`(J ��`HJ0��`hJp��a�J��a�J@��aK��a8K��aXK(��bxK�+��b�K�8��c�KP=��c�K`J��d�KpO��eL0\��e8L l��gXL�~��hxL`��i�L`��k�L��l�LPچ��m�L��oM��p8Mp��rXM�.��s�M`>��t�M�I��t�M@M��t�M�X��u(N]��uHNpi��vhN�x��w�N��x�NP��z�N��{�N྇��|OP؇��}(O0��~HO��hO� ��O��͂�O�.��O�9��O`=��X�H��(X�X��HX�_��hXo��X�{��؇�X ��X��XЛ��Y��(Y0��HY��hY��Yp��Y��YP��Y��Z��(Z� ��ݐHZ��БhZ��Z*��Z�.��Z�>��Z�M��[ ]��([�h��H[�o��h[`|��[��[@��Θ�[��[��\��(\��H\��h\��\��\��ޞ�\��џ�\��]p��(]��ڡH]P��h]�*��Ȣ�]0.��]�9��ˣ�]�I��Ȥ�]�X��^�_��(^�l��H^�o��h^�|��^��^Л��^��˪�^`��_��(_p��ŬH_P��h_0��_��_`��_��_� ��h��α(h��Hh,��hhP<��hL��h�[��ٵ�hp_��h o��i@|��(i ��Hi��ݸhi��i0��i��i��i��j��(j��Hj ��hj��j� ��j@��j�(��j0.��k@:��(kP>��HkPJ��hk�Y��k�h��k0o��k |��k ��l@��(l��Hl0��hl��l��l��lp��l��m��(m��Hm��hm��mp��m`��m`+��m�8��n�?��(npN��Hn ]��hn i��n�o��n�|��n��nЌ��oИ��(oP��Ho��ho��o��o��o��o��x��(x`��Hx`��hx��x��x,��x8��x<��y�K��(y@Z��Hy�h��hy�m��y�y��y��y@��y��z��(z��HzP��hz@��z`�]��threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_52_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��U( ��EDT��*�Re�H)#�E!�-��B��S�0�� A9@S��L@W��~ʪ�@[��@�WE0� �T@c�E��,2��(BhQZG�X��"?��[�)RE�"�Ȓ"XB�E�!��/��E{E@�/��(�Q`H��F�ЈateR�L#ň�"��5F��"��IF��ʨ2��\F�Q2EHF3ڌ�):��ge��2\Q$Z�(�"�W(�V�+�*�P+�k�A�� pE\��ke��.K�2\qW($[��K��W��V�+��p+��B�� !��pEb��oe��1R�2\�X5$\�H�"�WPFW�+.��+��%Cʕ��!��`�(b�*�E��]W�ȡ�@Z��!��Z��E��;��E��:�ȼ"tw�g(�zz�zE��/j��F��(�"�g�W��ct_�{��3�/�a�ˈ?C��/��?��?#�CN�,�@� ��,�k��;� �qE\w�q��`!��b\qAX�+�2��W�Da1��,,��qaX�+��;� qEb@� ��B\��_~!�q�f��/Db!��r�D,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8no�P�� @��av@�1�7��P�av@h��Aw�B"�� F.�� @��\0��`�2�`y7��|r�8�n ��q��@;��ແv@��w�R-�� ].�� @�ڿ\0��`D�A.G� @�j� �C� ��h��1z�~5�� v@x�q��@;�x�8�n �P;��7��Y@��`��- �� @��@�� }� �qz�FPp�@��v@�(��q��@;��\�8To �P{ �`�7�hOW0��`��+G� @��]�� z��Yp〽�v@�-��q��@;�\�8do �P��`�7�h�W0��`��+G� @�:�� "��-@��r� �H�9xC�� `��%0D��!��"��Y@�`u�, �H�:`C$&��!R��@� n�� `�t�C@��w��!��0D��2 �Hq<hf@�i� �"��0h@��K �"��C��8 `��D@�v{��!|�0DR�� Ha>px@��χ+�"��C$x�Q `�T�6 @�~~�J��!��\)0D��: �H�?p�@�� `��C�Y$�`�"�$��C��$�W `��@�ȓ�f��!R�V,0D�Q� �HoJ�@��N �"�)��C$Y%�^``�d�� @�J�x��!��8P0D"[2�HtKD��t��@"�.!� C$�%$y@`��T,�Z��!R?P0Drc"��HLH"��)� �$@"�31� C${&��@`��,�z�hb��!�ZM0D�k¹ �H�M@:@�鶉�'"�7�C��& � `�t�T�r�x��!��Z0D�sr�H�NDq@�)׉�."�;!�C$~'$� `��B��!R�i0D}B: �H�OL�@��5"�?A�C$,�� `�$�Ee�r��!+o0D�� H�X\�@��%��<"�e!�C$�,&� `�T��r��!RzC|0Db�B��H�Y@�@��E�!>�"�h!�C�1-�� `��3��`��!R�H|0D��H�Z0�@�i]g>"�k��C��-�� `�Բ�3�v��!��|0D��"��H�[L�@�i|g>"�o�C�.�� `��4�~�p��!R�|0D��H�\@�@�i��i>"�sQ�C��.�� `�d�E5�ֺ`��!�]L}0D킩�H�]0�@�i��>"�w��C��.�� `�T�s�f�X��!��J~0DB�b��Hs^$�@��ϋe?"�z��C�X/�� `��E�@�̽`��!��~0D��H&_$�@��$?"�|��C$�/�� `��ű@��H��!R��~0D �"��H�_ �@�I��d?"�~q�C$�/�� `��@��8��!��E0D��H&h�@��?�"ɠq�C�4� `�ԃ��@��!R�0D�C#��H{h�@�) �?"��Q�C$W4� `�d�1"@�t�@J��!�8J�0DbG9�H�h$'@��?"��A�C$�4�� `��@�n�HF��!Xň0D*KCQ�Hoi*@��.��?"��A�C��4 `��"@�`�PJ��!�nF�0DO��H�i�@��=��?"ѧA�C$�4�� `��F�@�J�@��!R��0D�Q��H9j�@��G �?"��Q�C$55�� `�$�F�@��PF��!R�ƈ0D:U��H�j�@�IV �?"�Q�C�_5�� `�t�F�@��H��!R�F~0D�Wc)�H&k%@��e �E"嬁�C��5�� `��F�@��H��!��0D�[��H�k�@�Iu��?"ͮa�C��5�� `�$��@��H��!R��0D�_#��H l�@�Ʉ eD"��a�C$6�� `��F�@��8��!��0Drc��Hvl�@�i� �?"��q�C$V6�� `�d��@�z�hF��!�:Ɉ0D�g��H!m(#@�i� gD"崡�C��6 `��#@��H��!h�0D2mC��H�m2@�鶍EF"��C$�6�� `�4߆�@��@��!R��0D2q#��H/n2@�� EF"��q�C$37�� `�4��@��HF��!��ň0DZu#��H�n4@� ׍dD"��Q�C$s7� `��F1"@��H��!R��0Dy#��H,o�@�I��?"��a�C��7�� `��@��H��!R��0D�{��H�o$6@� � �F"��q�C��7�� `�4��@��0��!��0D��H&x8@�)�?"��1�C$<�� `��@�v�8p��!��0D*�C��Hwx8�@��?"��C�r<�� `��@��X��!RH�0DJ��H8y,�@�i,��?"��C$�<�� `��@�J�ht��!�l�0D�ͣ��H�y�@�i=$>"��C��<0 `�4�G&@�P�@��!��0D��H>z `@��LL"�遀 C$;=0 `��&�L�@F��!��0D�Ճ� �H�z >@�I\DG"��A�C$w=�� `�d��@��!��0DZ��H3{`@�)gL"�큀 C��=0 `��&��@��!R��0DJ��H�{ <@�)w�G�"��C��= `�Ծ�#&��`��! ��0D�� Hc|<b@�I��GL"�� C$Y>1 `��G$&��h��!R?�0Dr�! �H;}<b@�I�FL�"�� C��>1 `��#&��`��!{��0D�� H'~<b@��FL�"�� C�8?1 `��#&�X�`��!��0D:�� H�~8b@�)�EL�"�� C$�?1 `�T��#&��`��!��0Dz�� H�@C��?��!��@��f `�4��c&��q��!"��0D E�1 �H��@h@� M�"�#r� C�zD 6 `��ȡ&0�B)��!�I"��0D�I� �Hf� p@��-N�"�%�� C$�D09 `�d�H�#@�~Ix��!z" �0D�O$� �H��`r@�IF��G�"�(��C$2E `�ԦH�#@��!��" �0D�U$� �H��$<@�)]��G�"�+�� C$�E�� `��ȳ@��y��!R�"�~0Db]��H�@�@�ih?�"�0�C�F�� `��g'�Nq��!R-#�0Dg�� H�8x@��N"�4�� C$�F< `��ȃ'@�Ty��!Rn#��0D�o�� H#�8|@�)��O�"�8�� C�6G<= `��'@�ty��!R�#�0D�w�� H'�x~@��O"�>�@��L"�? C��G3 `�䂉c&@�x0y��!�&��0D��1�H��4f@��L"�c C�wL3 `��Ic&@�V2y��!�N&��0DB��1�Hw�8f@��4��L"�f� C$�L3 `��Ic&@��3y��!�&��0D��D�Hd� l@��MKT "�j2� C�ZM(5�`��I�&@��5��!R�&�0D��$��Hn�d�@��t�("�n�� C��M�8 `�D� F*@4�x8y��!�'��0D��3�H�|�@�)��00�Hb�x�@� ��"�v��C$�N�Z `�� W��<Y��!�'�|�0Dj�dB�H��X��MU"�|�� C$�OT `��Ʌ*@<�`?��!�'S�0D��H#�@�@�i�)>"��I��Pa��!�(*�|0DzE��H��<�@��i>"��"�C��T�� `�$�J4��Ra^��!�_*ϫ0DrMEz�H�D�@�)>&V"�� C�U�X `��+��T�b��!R�*R�0DZU%��H��D�@��]fV"�� C$�U�Y `��J4+��Vab��!��*L�0D]��H��<�@�)|)V"ͯ"� C�V�X `�d�J+��X�j��!�+׭0D�ee��H�X�@�I��V"�� C$�V�Z `��U+�n[�j��!�{+S�0D q�� H/�8�@��+W"͹�� C�UW�\ `�$��+��]�z��!�+V�0D�y��Ht�`�@��W"��r� C��W�^ `�$��+��p�~��!�.د0D��H�L�@��W"��b� C$�\�^ `�ė�+�hs�z��!{.W�0D��H:�,�@��L�%\"�鲌C�=]�q `�d��b.D�buY �H��,�@��\(]"��C��]�w `��K$/��vy��!�.�0D:݅��H��,��|��\�"��@��wY��!� /��0D��% �Hi�H$@�i��d"��C�^^�� `��΋3�FzYd��!�K/K�0D��%��Hj�D;@�i��g"��C��^�p `�T�3.D��{a��!R/��0DJ�"��C�0_�t `�$�K�.�F}��!R�/��0D:��H��@�@��j_"��C$�_s@`�D�0D�� H��H�@��)d"��2�C�d�� `�d�L�2�ΐ�Z��!2�0DzE��H��05@�� g"�##�C��d&� `��.�Ԓa��!�]2̸0DM�"�H��0�@�I7 `�d�.�䓉��!҈2�0D�QFz�Hf�H�@� O�^"�*�C$\e�} `�Ԯ3.H��a� �!R�2@�)f&\"�,#�C��e�� `��D2�h��N��!��2��0D�_&j�H)�\0@� �Ff"�1��C$<f$� `��ʌ�3�x��|��!�;3L�0D�g� �H'�0�@�i�f\�"�4ØC��f��!�]3L�0Dm&J�H��D�@��]"�7#�C�g�z `��/�Ҝ��!��3̸ 0DRu�1$�H��0C$qg�p `��/�@��B��!��3�0D*{f:�Hx�D*@�)��e"�>s�C��g� `��S3�R��p��!�6��0Dj�f��H��0�@��&\"�bÌC$rl�q `�� c.D��a �H(�0�@��&�(]"�e�C$�l�w `��$/�t��!�z6�0D�Ϧ��H/�0��iG�\�"�i�@��a��!ҟ6��0D��F �H��L$@��^��d"�l�C$�m�� `��3�D�ad��!��6L�0D��F��H��H;@��g"�pÄC�n�p `�� 3.D��a��!R(7��0Dj冁"�rÄC�rn�t `�d�M�.�N��!RN7��0DZ��H{�@�@�i��j_"�wÌC$�ns@`�� 0D"� �H0�H�@�I�)d"�y3�C�So�� `��M�2�ֽ�Z��!�7�0D��H?�05@�i� g"�}#�C��o&� `�$� 6�ܿa��!��7��00D2A�&�H2�0d��l"��À C$;t$� `��6�tљ��!R;:��0DI'��H1�Dr@�I,�n"�È C$�t��`��c6�x�a��!z:��0D�Og��H.�Pz@�IL�o"٩#� C$Tu$� `��&:��aH��!��:��0DRYgB�H=�L�@� n�u"��À C$�u� `�Ļ#6L��a�� !�:�00Da�1�H,�0`@� � m"��#� C$>v&� `��6��ى��!�<;��0D"iǢ�H:�0b�ɬ�l�"ɵØ C$�v� `��d7�zۙ��!|;��0D"q��H8�H~@�I� t"񹃉C$Zw� `��c:��ݙP��!ҿ;��0D�y�b�Hf�0`@�I�l"��È C$�w�``�d�C6L��a��!��;�0D�GB�H,�Hj@��m"��3� C$R|"� `�T�O$7��!?>��@0DBɇ!(�H4�0f@� ,l"��3� C��|&� `�$��7��!�~>��0Db�G�H>�`�@��N�t"��ØC$W}&� `��Ϥ:��X��!�>�0D�ه�H`�0b��mFl�"��Ð C$�}� `��6�x��!��>��0D�gb�H3�Ln@�Ɍ�n"�� C$T~,� `��#6P��a� �!�=?��0D��H.�Lv@�)�� o"��C� C$�~(� `��ۏ�7��@��!�?��0D��!�Hd�0�@� Ο u"��3�C$[(� `��5�@��!��?�|0Dj��$6��$2�(L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B��1��G�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align&0��v#��#��#� #�ڍ ��8� ��9� ��9� ��:� l7��0��0��0�0�0��Y��[��\��^��_��Bi��j��m��m��x��x��z��|��v#��#��#��#�A#�Q#��#��#�R#�R#�"R#�2R#�BR#�RR#�bR#��R#"�R#%�R#(�R#+�R#.S#1S#9CS#<SS#?cS#��S#��S#�V#�V#�"V#�2V#�BV#��V#"�V#%�V#(�V#+�V#.�V#1�V#9#W#<3W#?CW#��W#��W#��W#��W#�Z#�Z#�"Z#�rZ#"�Z#%�Z#(�Z#+�Z#.�Z#1�Z#9[#<[#?#[#�[#��[#��[#��[#��[#��[#�^#�R^#"c^#%s^#(�^#+�^#.�^#1�^#9�^#<�^#?_#�_#�_#�_#�_#��_#��_#��_#�2b#"Cb#%Sb#(cb#+sb#.�b#1�b#9�b#<�b#?�b#�bc#�rc#��c#�c#��c#��c#��c#�f#"#f#%3f#(Cf#+Sf#.cf#1sf#9�f#<�f#?�f#iCg#�ۍ �M�� p7��6v6��6w6��B�� w#u�g#w�g#yj#{j#}#j#3j#�Cj#�Sj#��j#��j#��j#i#k#k3k#mCk#oSk#qck#ssk#u�k#w�k#y�k#{�k#}n#n#�#n#�3n#�cn#�sn#��n#io#ko#m#o#o3o#qCo#sSo#uco#w�o#y�o#{�o#}�o#�o#�r#�r#�Cr#�Sr#�cr#i�r#k�r#ms#os#q#s#s3s#uCs#w�s#y�s#{�s#}�s#�s#��s#��s#�#v#�3v#�Cv3�0�@�C3Fp�0 �1Ð��3K��043��03P�0HA4�0�D3U�0XA4�p�D3Y�0hA4ð�D3]�0xA4��`0��0�A3c��`0�P�0�A3g��C��C��C��C��C��C��(�C(��(��(�)�C)��)��)�*�C*��*��*�+�C+��+��+�,�C,��,��,�-�C-��-��-�.�C.��.��.�/�C/��/��/�8�C8��8��8�9�C9��9��9�:�C:��:��:�;�C;��;��;�<�C<��<��<�=�C=��=��=�>�C>��>��>�?�C?��?��?�H�CH��H��H�I�CI��I��I�J�CJ��J��J�K�CK��K/1�AL�0�D33��CM41�`�M�0�Dp39��N11ðAL�0�D@3=��O71��M�0�E@3a��X71�0�K�0�E3e��Y41�pM�0�Ep3i��Z/1ðAL�0�E3m��[41��M�0�Ep3q��\71�0�M�0�E�3u��]11�pM�0�E@3y��^71ð�K�0�E3}��_41��M�0�Fp3��h41�0�M�0�Fp3��i11�pAL�0�F@3��j71ð�M�0�F�3��k11��M�0�F@3��l71�0M�0�Fp3�� b�h�b�h�h h�h�b�h�h�h�h�h�h�h h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�b�b�h h�b�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h��&(7;��49�2��7�Q��H��ή�%��.�.��mnհ��Ѝ�� 0��@��P��`��p��Џ��!�1�A�Q�a�q��ё��0�1!�21�3A�4Q�5a�6q�7��8��9��:��;��<ѓ=�>�?�P�Q!�R1�SA�TQ�Ua�Vq�W��X��Y��Z��[��\ѕ]�^�_�p�q!�r1�sA�tQ�ua�vq�w��x��y��z��{��|ї}�~��!��1��A��Q��a��q��љ�ᙞ�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d�}\�}��4Ma A,$�� d�}\�}��4Ma A,T�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,�� 6C�A8 �}\�}`~��4M�}��4Ma A,D�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,t�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� d�}\�}��4Ma A,d�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,D�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,t�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,4�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,4�� 6C�A8 �}\�}`�}��4M~��4Ma A,d�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� d�}\�}��4Ma A,�� d�}\�}��4Ma A,$�� 6C�A8 �}\�}`�}��4M~��4Ma A,T�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma A,T�� d�}\�}��4Ma A,�� 6C�A8 �}\�}`�}��4M~��4Ma A,�� 6D0�B�@ �}\�}`�}d~��4M~��4M&~��4Ma A,�� A�6E0��lCP�~��4M6~��4M�}X�}\�}`�}dF~��4M&~��4Ma Ĉ�� G� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` ��kBp�p r2b` �nBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a Ĉ�� ć� �pC`f0��,�0K0TD�B�@�}0�c�`�A`��4�y�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �kBp�p r2b` ��kBp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1�~`NSٽ�$@#��ұ�� $@#��Twm��4Ս[�� 0��,6o��v`NS��T8m��4n��4�m��4Un��4o��4��T5n��4U��AX�A��!�n��<wn��4�M��!8Meۦ~@��X��!8Me�a �� ć� �� ǬB�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ�ұ�� T7n��4խ��!8Mu�~0�`��4��DF~@̀��A!8Me[�A!8M��A!Q}��A!8M�[�!8Me��4U]��!8MU�`��ܷ��Txo��4U��AX�A��A ��!�n��,n��!˂�`NS巡@3Դ��T�m��4��!�a �� ć/ �� /B�@�}�}�}�}`��,>Ra �� G/ A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��Nd��4��!�r�F~@̀��A!8Me��A!8M�[�A!8M�[�!8Me��4U]��!8MU��~`NSպ��T5o��tD�m�`�A`��`��tD��,�m��5m��4�}��!�a �� ǫ �� B�@�}�}�}~P��}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|��\p�@�@T� WC��5�~`NSٽ��,�~`NSٶ�ұ�� T7n��4խ��!8Mu�~0�`��4��D6~PNS�v`NSὕ��T��@3�~PHGT�~@NS��A!8M�[��!8MU׶~`NSո��T�m��4U��AX�A��A ��!�n��,�}��!˂�f`��༡@3Դ��T6a �� < �� G<B�@�}�}�}�}`��,>Ra �� ; A�@�}�}�}@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|��\p�@�@T� WC��5`NSټ�ұ�� T7m��4�m[��!8Mu�`��ػ�@3��!8M�߆`NS��T�n��,˭�4n��4�m��4Un��4o��A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o�`�A`��`��tD��,�m��5m��4�}��!�a Ĉ�� ć� �� Ǭ�pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a Ĉ�� ǫ �� pC�f0��,�0K0TD B�@ &~@N3X�� 0�q�u�y��,�}�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|��\p��7��`��7��#@�@T@WC��A��0H�b��tD��~0�`��4�m��4Un��4o��,�}��!8Me��4U]��!8MU��}�}V`��y��,8o��5m��4�m��!�b��4��!8Me�&`NSݸ��T�n��<F`NSݼ��T8n��!8M��V~@̀��A!Q}[��!8M��`NS帝��`NS�"��T� a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�R d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��,��[�A!8M��A!8Mś�A!Q}[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z d�}�}�}�}0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\@�@TWC��<`NSټ�ұ�� T�m��4Ս[��!8Mu�v`NSݼ��Two��`NS��T�n��4n��4�m��4Un��4o��tD�m��4��Tum��4U�[��!8M��`NS�q�u�y��`��๹�0��Aý�A4CM��!8Me��~`NSٸ��Tvn��4��!�a A ��A|�S d�}�}�}�}0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� ć� �w�E�9� \p\@�@��A|��\p\@�@TWC��<F`NSټ�ұ�� T7m��4�m��!8Muߖ`NSݹ�@3��!8M��`NS��T8n��4n��4�m��4Un��4o��tD�m��,��[�!8Me��4U]��!8MU��~`NSպ��T5o��4U��AX�A��A ��!�n�0fP�p׆~P�P�`NSٴ��T�m��,��T�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!Q}��A!8M��!8Me��4U]��!8MU�~PNS��`NS�q�u�y��`��"��Tym��4U��"��Tyo�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a � ��A|�Z 7x`��@@� d ~0�c�`�A`��`�0&~@N3�!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� Ǹ �w�E�9� \p\��A|��\p\�7��7��@ #@�@T�WC��H&`NSټ "$@#��"ұ�� $@#��T�m��4Ս��!8Mu��0�c��4��[��!8Mu��P̓�~0�`��4^��!8M��V~@̀�A!8Me��A!8M��A!8M��A!Q}��!8Me��4U]��!8MU��`NS��T8o�`�A`��`��4U��!�n��,�� !8M��~0�`�0�n��5m��Y��!8Me��~`NSٸ��Tvn��4��!�a �� ć� �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� ć/ �p1��Q�D8~0�c�`�A`��T��T�}`��a �� ǫ �p1��Q�D8 ~0�c�`�A`��T��T�}@��a �� < �p1��Q�D8~0�c�`�A`��4�n��T�}`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��kBp�p r2b` �!nBp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��kBp�p q1b` �!nBp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �kBp�p r2b` ��kBp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<`NSٽ��ұ�� Twm��4�}[��!8Mu�60�c��4��!8Mu��~`��ؼ�@3��!8M�Ӗ`NS��T�m��Y�4n��4�m��4Un��4o��4�yOD��X��!8MU�~`NSպ��T8o�`�A`��ÿ8v~@HG��!8M��~`NSٴ��T�m��,��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �kBp�p q1b` ��kBp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<&`NSٽ�ұ�� T7m��4�]��!8Mu�v`NSݺ��X��!8Mu��~`��ظ��Two��`NS��T8n��Y�4n��4�m��4Un��4o��4��T5m��4Um��D��T�n��4U��AX�A��D��/��P�s��~`NSٴ��,`NSٹ��T�a ��A� ��}\�}@��a ��A� 쇙��}\�}@��a ��A� ��}\�}@��a ��A� 䇘��}\�}@��a ��A� ��}\�}@��a ��A� 쇏��}\�}@��a ��A� ��}\�}@��a ��A� 쇎��}\�}@��a ��A� ��}\�}@��a ��A� 䇍��}\�}@��a ��A� ��}\�}@��a ��A� 쇌��}\�}@��a ��A� ��}�}@��a ��A� ��} �u�,a ��A� �G��} �u�,a ��A� �Ȋ��} �u�,a ��A� �G��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a ��A� ��} �u�,a �� Ȉ A�@�}�}�}@��a �� H� A�@�}�}�}@��a �� A�@�}�}�}@��a ��A� ��} �u�,a ��A� ��~��} �u�,a �� H~ A�@�}�}�}@��a �� } A�@�}�}�}@��a ��A� �}��} �u�,a �� | A�@�}�}�}@��a �� H| A�@�}�}�}@��a ��A� ��{��} �u�,a ��A� �{��} �u�,a ��A� ��z��} �u�,a ��A� �z��} �u�,a ��A� ��y��} �u�,a ��A� �y��} �u�,a ��A� ��x��} �u�,a �� Hx A�@�}�}�}@��a ��A� ��o��} �u�,a ��A� �o��} �u�,a ��A� ��n��}�}@��a ��A� �n��}�}@��a �� m A�@�}�}�}@��a �� Hm A�@�}�}�}@��a ��A� ��l��}�}@��a ��A� �l��}�}@��a ��A� ��k��}�}@��a ��A� �k��}�}@��a ��A� ��j��}�}@��a ��A� �j��}�}@��a ��A� ��i��}�}@��a �� Hi A�@�}�}�}@��a ��A� ��h��}�}@��a ��A� �h��}�}@��a ��A� ��_��}�}@��a ��A� �_��}�}@��a ��A� ��^��}�}@��a �� H^ A�@�}�}�}@��a �� ] A�@�}�}�}@��a �� H] A�@�}�}�}@��a �� ] B�A8�}�}�}0@�c��,a ��A� �\��}�}@��a ��A� Ȉ[��}�}@��a ��A� �[��}�}@��a ��A� ��Z��}�}@��a ��A� �Z��}�}@��a ��A� ȈY��}�}@��a ��A� �Y��}�}@��a ��A� ��X��}�}@��a �� HX A�@�}�}�}@��a �� X B�A8�}�}�}�}@��a �� HO A�@�}�}�}@��a �� N A�@�}�}�}@��a ��A� �N��}�}@��a ��A� ��M��}�}@��a ��A� �M��}�}@��a ��A� ��L��}�}@��a ��A� �L��}�}@��a �� L B��@�}�}@H��!<Sa �� ؈K B��@�}�}@H��!<Sa ��A� ��J��}�}@��a ��A� �J��}�}@��a �� I A�@�}�}4�}@��a ��A� �I��}�}@��a ��A� ��H��}�}@��a �� HH A�@�}�}4�}@��a ��A� ��?��}�}@��a ��A� �?��}�}@��a ��A� ��>��}�}@��a ��A� �>��}�}@��a ��A� ��=��}�}@��a ��A� �=��}�}@��a ��A� ��<��}�}@��a �� < A�@�}�}�}@NS��!�a � �� <B�@�}�}�}@NS��!�a �� ; A�@�}�}�}@��a �� ; A�@�}�}�}@��a "D,��A|Ě\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|ȓ\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� : A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� I: A�@�}�}�}@��a �� N A�@�}�}�}@��a �� 8 A�@�}�}�}@��a � �� H8B�@�}�}�}@NS��!�a �� H8 A�@�}�}�}@NS��!�a �� G/ A�@�}�}�}@��a �� G/ A�@�}�}�}@��a "D,��A|К\`��#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa "D,��A|��\`@�#�`&�t��@�A��A86~`��Է��T��T�}��T�}�}~@��AS,�Oa �� I. A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� - A�@�}�}�}@��a �� HO A�@�}�}�}@��a �� + A�@�}�}�}@��a � �� +B�@�}�}�}@NS��!�a �� + A�@�}�}�}@NS��!�a �� * A�@�}0T�b�0!I�}@��a �� * A�@�}0T�b�0!I�}@��a ��A#�` &�4��}0T�b�0!I�}04�a��,a ��A#�` &�4��}0T�b�0!I�}04�a��,a �� ) A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� ( A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� ,g4A�@�}0T�b�0!I�}p�a��,a �� ,e4A�@�}0T�b�0!I�}p�a��,a �� A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� B�A8�}0T�b�0!I�}p�a��,a �ȈA� � � ��@�@ �}0�c��4�u��T~PNS�~PNS��}@��a �ȈA� �� @�@ �}0�c��4�u��T~PNS�~PNS��}@��a ;J,@ �� a��pC�`0�0�� %�1K�7�2F0bp p�C��#�`'?�,�,�1PAD!��@y(�p�$��A8V~`��Խ��A!0OM��!8̀Ӷ~`N3൝�,7m��,wo��<vm��4�q�u�y��,a 9J,@ �� ܐQ�$� E��,�@#�`�#8��p�Q��,Ca#�`�#9�$�ǈ�� H-�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~`N3ؼ��6~@̳��!8�`�f~`��ص��AX�A��!�b�V~P�S�&~@��a 9J,@ �� ȐQ�$� E��,�@#�`�#3��p�Q��,Ca#�`�#4�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛��!8�`�&~@��a 9J,@ �� Q�$� E��,�@#�`�#.��p�Q��,Ca#�`�#/�$�ǈ�� -�,�1PAD!��@y(�p�$��A8F~`��Էy�3�~@N3��!0�n��,�][�� 0�q�u��,�n��<un��4�͛�!�b��4��a 5J,@� ��,C #�`�&D�$� ��,�@#�`�'d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�!D�$� ��,�@#�`�"d� � F��~�2Kp��~�P2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 5J,@� ��,C #�`�'�C�$� ��,�@#�`�'�c� � F��2Kp��O2KpTD!��@��1K�TP ��}0�c�`��4�u�3v~`N3Զ��m��,Kn��<��,��~@��X��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |R?�,�1bP |R?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a 7J,@ �� O1K�7x��1K�7�2F0bP |?�,�1bP |?�,�1PAD!��@y(�p�$��A8~0�c�`��4�u�3�~`N3Զ��m��,Kn��<��,��~@��!�R�a � �X�@�}�}@0�@��!8Ma � �T�@�}�}@0�@��!8Ma � �`�@�}�}@0�@��!8Ma �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �� I�A�@�}0T�b�0!I�}@��a �� A�@�}0T�b�0!I�}@��a �ȈA� �I��}@NS��D�`��,a �ȈA� ��}@NS��D�`��,a �ȈA� �I�� @�@ �}0�c��4��T��T�qOD�}@��a �ȈA� �� @�@ �}0�c��4��T��T�qOD�}@��a ��A� � ��}0<Sa ��A� ��}0<Sa ��A� � ��}0<Sa ��A� ��}0<Sa ��d�@�}<a ��d�@�}<a ��d�@�}<a ��d�@�}<a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a ��p�@�}pD� �a �� ̇�B�A8�}`8E!�q )�}@�Da ��A� ��} a �� J��}0�d�2a ��A� ��}@0��a �pCPa0bp ��6�� Ap�A8�}@<1��a A,@ � �� B0��}@<1��a �pCPa0bp ��6C� Ap�A8�}@<1��a A,@� �� B0��}@<1��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d �}PX�0F~ 8��ұX��!EB 4�q�� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| d�}PX�0&~PHT�#�}pH�P a�CDF~@8��A!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ �}PX�0~PHT�#�}0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u�L��a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��@,`��ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d �}PX�0F~ 8��ұ��P�u �& )&~0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d�}PX�0&~PHT�#�}0T@d��4� H��N�#6~PH��#a &A,@�A5��<��9��9��A��6!P�\Ď�� d �}PX�0F~ 8��ұX��!EB 4�q�� Qa #A,@Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@�}PX�0&~0H�b�CD�}P�2~0HTa %A,@�A5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u�L��a &A,@�A5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d �}PX�0F~ 8��ұ��P�u �& )&~0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�}pH�P a��m��t,>b��HLDH�2Ֆ~@8��!Q��Ԇ~PH��#V~��L�}0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~PHT�#�~`H��#��(��D�D-Sm��D>R��A!��AHLDH�2��P�u�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��f~��L��ұ��m��Hm��t,>b� 1!Q�d�CD�}P�2a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a UA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��Hm��t,>b� 1!Q�d�CD�}�,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m�� %��~�HT�#>R�"��m� 1!Q�T��!Q��ұ��X�AHLDH�2��P��N�#>a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@d�}PX�0�~�HT�#>R��"��m� 1!Q�T��!Q��$��D�D-�qY�A1��!��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��X�AHLDH�2��!��qY�A1�X��D>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPd�}PX�0�~p8��m��D>�#�� ұ��f~��L��!��$��D�D-�qY�� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d��D>R��N�#>b��t,>�#V~��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#��uR$�@C�A"��X��D>�#>b� 1!Q��#>b�CDv~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0f~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`��D>R��ұ��HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`P;�1�� B`��| ��@~p#G�`0��#F�`p;J@d�}PX�0�~�HT�#��$ұ��4��D�D-��qY�A1��A"Q��"��>�#V~��L��a WA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�~`8�� H�#�~pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��@,`��D>�#>b��#>b��t,>b��HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}�,@C��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l d�\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`P;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`p;LPd�}PX�0�~�HT�#�� N�#>�#�~�H��#>�#V~��L>�#�}0T@d��4� H��$��$N�#��(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұ��!EB 4�u�� Q�q )a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#�}pH�P a��ұ��u��a A,@�&!P�܈A� �J� �a� �A8 �}PX�06~0H�b�CD�}P�2�}@0�� Qa A,@�&!P�܈A� �J� �a� �A8�}PX�0&~PHT�#6~PH��#�}0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@�}PX�0F~ 8��ұX��P�y�L��a A,@�&!P�\Î$��L��&@�@�}PX�0&~PHT�#F~@8��A!��X��P�y�L�q )a A,@�O6!P\Ď$��B$H0��}PX�0V~ 8��ұX��P�q )�}�,@C�� Qa A,@�O6!P\Ď$��B$H0�� }PX�06~PHT�#V~@8��A!��X��P�q )�}�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d�}PX�0F~ 8��ұ��!EB 4��,�m�CD&~0HT`�H��,a A,@�&!P�܈A� ��# �`p+.��!�&��A� �� }PX�0v~`��Է�ұX��P�y�L�q )&~0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d�}PX�0F~ 8��ұ��!�R��}0T@d��@,`�H��X�!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8�}PX�0V~ 8��ұX��!�R��}0T@d�H�y �& )6~0HT`��,a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�yR$�@CX��D>b� 1!Q�d��D>R[��Pa MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a FA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��t,>b�H��Hm� 1!Q�d�CD�}P�2a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��,��D�D-��!�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~PHT�#�~`H��#��(��D�D-Sm��m��t,>b�H��Hm��$��D�D-�u��A1�a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a JA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~PHT�#�~P8��~`H��#��(��D�D-Sm��t,>b�H�y �& )�~`HT�#��$��D�D-��N�#�}0T@a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a MA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�yR$�@C�"Q��Hm��t,>�#��$��D�D-�u��!��a GA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ~�E�@�}PX�0f~��L��b��D>�#�� ұ��V~��L�}0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��i7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d�}PX�0f~��L��b��D>�#�� ~�H��#>b[�AHLDH�2Y��P�y�L��ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�vsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@�}PX�0v~��L��b��4� H��N�#>b��~�HT�#>R��"��m� 1!Q�d�CD�~pH��#>a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�}pH�P a��D>�#�~p8��>b��t,>b��0��D�D-S�#�}0T@d�H��N�#>b��t,>�#V~��L>b��D>R�a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`p;N`d�}PX�0�~�HT�#��yR$�@C�A"��X�AHLDH�2��X��D>�#>b�CD�}@0��A"Q��A�D>r��X��"��>�#f~��L��a HA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0v~pHT�#>b��t,>b��0��D�D-S�#�}0T@d��@,`�H��ұ��X��HLDH�2�� H�#a JA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��$��L�K0��0!P(܈A� ܊�D�� 0b`T ��E�@�}PX�0�~�HT�#��$ұ��4��D�D-��u��A1��b��D>�#>b��t,>b��HLDH�2�>�#a MA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0v~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}P�2�}@0��X��D>�#�~pH��#>b� 1!Q��#�~�HT�#��a PA,@�Y7�P��`t�@p�d��`@r�G��<�;vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`P;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`p;N`d�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d��@,`�H��$��$N�#��(ұ��#>b� 1!Q�T��a MA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~pHT�#>b��#�~�H��#��HLDH�2�>b�CD�}@0�� a�b��X��!��,��D�D-��X�"Q��>a PA,@�O�usy��`n v@`�dl@q�ld {p�<�A��a7WP��`tdl�@q�d ��F��#E�`P;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@�}PX�0�~�HT�#��$ұ��4��D�D-�� N�#>�#�}0T@d�H�y �& )�~�HT�#>�#�~�8��>�#�~�H��#��8��D�D-S�#>a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8MU�&~@NS�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�6~��L��T5n��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pn0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pn0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��pnPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��pnPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��pn`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��pn`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0n0q``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0n0v@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,T�q7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,�}7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8�}0<Sd��4a A,t�\ah�@np��0nPq``@p�`�d��D��#E�``.��A8�}0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@�}0<Sa A,$�\ah�@np��0nPv@`@p�`�d��D��#E�`.��A8�}0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\ah�@np��0n`s `@p�`�d��D��#E�` 3��A8�}0<Sa A,d�\ah�@np��0n`v@`@p�`�d��D��#E�`P3��A8�}0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa .A,�� 8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,�}7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_�_�H#�d��4�m��A��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4M6~��L��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��_�O�H#�d��4UM[�A��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L�q��HLDH�2�F~@NS��HLDH�2�V~`NSո��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�F~��4Mf~��4M&~��L��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,T�q7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,�}7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�;�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��4�m��A��>�HD4Y�A��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`@>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�``>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�&~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8�}0<Sd��4a A,t�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�``.��A8�}0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@�}0<Sa A,$�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`.��A8�}0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �~��L~��L�q��!8Me��4UM�A��>�HD4Y�A��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!�~��4M~��L��(��D�D-�n� 1!Q��[��!8MU��}0<Sd��4U��AHLDH�2�A��>�HD4�!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`0?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`P?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`p?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,4�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�` 3��A8�}0<Sa A,d�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`P3��A8�}0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a EA,��8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8~��L�q��HLDH�2�6~��L��4~!>�#�DD��4~!��#�DD��$��D�D-��4~!~�#�DD��4~!��#�DDa 2A,�� 8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�` B�0�� B0�P�lC �~��L~��L�q��A��>�HD4��A��>�HD4a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�``B�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`�BC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�` CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�`@B`8�}0<Sd��4�u�Pa A,T�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�`@6� `8�}0<Sd�E,a A,t�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�``6� `8�}0<Sd�E,a A,�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�`6� `8�}0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�`@C�!~0U�B~!~��d��4n��4�q��TE��O\S�u�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�``C�BB��@~0U�B~!��d�S,��>qM LF~PNS�V~PNS�6~@NS��!8MU��}0<Sd�E,a�S,��>qM L~0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`�C�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�`FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,T�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�`@:� `8�}0<Sd�E,a A,t�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�``:� `8�}0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�` FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�`@F�AB��@&~0U�B~!��d��3E�}PT�B�}0U�B~!>��d�S,��>qM L~0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�``FC!�}0<Sd�S,��>qM L~0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01��} �q�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01��} �q�c��t �q #2"�\�.�E�^�.��`�.��b�/86��d�/XF�f�0x� �h�0�� j�0�6�l�1�F�n�1��p�1�)�r�287-�t�2XG8�v�3x�:�x�3��=�z�3�7I�|�4�GL�~�4�7O��5@Z��58�\��5X�_��6x0k��6�@n��6�0y��7��|��8��98`��:XP��:x௃��;��;��ȃ��<�0؃��=�0郠�?��?8��@X��Ax��B��.��C��?��D��X��E��[��E�k��G8 ��Hh ��H��I� ��I(��Ih ��I��I� ��J��JH ��J��J�0��JЯ��KHP��K��K��K��KH��L�Ʉ��L��ʄ��L( ̄��LH(�̈́��L�(0τ��M�(�؄��M)@ڄ��MH)�ۄ��M�)@݄��M�)�ބ��M*P脢�NH*�鄦�N�*p넪�N�*0턮�N+��NH+0��O�+��O�+0��O,��OH,��O�,��P�,� ��P- ��PH-��P�-P��P�-��Q.`��QH.��Q�.`��Q�.��Q/��R(/`)��Rh/�-��S�/`:��S�/<��S�/�=��S�/@?��T(8 I��TH8�J��T�8@O��U�8�[��U�8`]��U�8_��U9�h��V89�j��Vx9`l��V�9�n��V�9`x��W:@z��WX: |��W�:~��W�:��X;��XX;��X�;��X�;��Y<��YX<`��Z�<ૅ��Z�=@��[8> ʅ��\�>م��]x? ��^�?��^�?P��^H0텂�^HH�_�H��_�H��_I��_HIp��_�I0 ��`�I� ��`J��`(J ��`HJ0��`hJp��a�J��a�J@��aK��a8K��aXK(��bxK�+��b�K�8��c�KP=��c�K`J��d�KpO��eL0\��e8L l��gXL�~��hxL`��i�L`��k�L��l�LPچ��m�L��oM��p8Mp��rXM�.��s�M`>��t�M�I��t�M@M��t�M�X��u(N]��uHNpi��vhN�x��w�N��x�NP��z�N��{�N྇��|OP؇��}(O0��~HO��hO� ��O��͂�O�.��O�9��O`=��X�H��(X�X��HX�_��hXo��X�{��؇�X ��X��XЛ��Y��(Y0��HY��hY��Yp��Y��YP��Y��Z��(Z� ��ݐHZ��БhZ��Z*��Z�.��Z�>��Z�M��[ ]��([�h��H[�o��h[`|��[��[@��Θ�[��[��\��(\��H\��h\��\��\��ޞ�\��џ�\��]p��(]��ڡH]P��h]�*��Ȣ�]0.��]�9��ˣ�]�I��Ȥ�]�X��^�_��(^�l��H^�o��h^�|��^��^Л��^��˪�^`��_��(_p��ŬH_P��h_0��_��_`��_��_� ��h��α(h��Hh,��hhP<��hL��h�[��ٵ�hp_��h o��i@|��(i ��Hi��ݸhi��i0��i��i��i��j��(j��Hj ��hj��j� ��j@��j�(��j0.��k@:��(kP>��HkPJ��hk�Y��k�h��k0o��k |��k ��l@��(l��Hl0��hl��l��l��lp��l��m��(m��Hm��hm��mp��m`��m`+��m�8��n�?��(npN��Hn ]��hn i��n�o��n�|��n��nЌ��oИ��(oP��Ho��ho��o��o��o��o��x��(x`��Hx`��hx��x��x,��x8��x<��y�K��(y@Z��Hy�h��hy�m��y�y��y��y@��y��z��(z��HzP��hz@��z`�]��threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_53_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU�_~��DU!�!ȪR@X�)H�R��/@\�ȫR�@`��4@d�� ,��w�EiQE(c-b��(Bn�[�G��"E��]TYRK�"�(�"��E�A��/B�hF�E�c0"�)B ��a�HI��L�� c��RR#Ȩ�"T#MV�_�ʊ@�Pt S�a2#�(�"$�uF��2\р$Z��"�W�u�V�+�P+�k�A�� pE\~�ke��.E�2\q�%$[��+��W�% �V�+C�p+��A�� !��pEb��oe��L�2\�)$\��"�W$�W�+�"��+�� Cʕ� Β!��p�g�sS�T0�@ �1�@F�"Xg]W��+�u��vE�n-��E��-�w�Q��q�w�Q��"�`�]�^�Ӌ�3��"X�W�� a4_�K��+�u��}E��1ʯ�IF��_��_f�!~��_� g��!�~F�� q�Z@� �H�B\�w�Ÿ��;��r�W�a1��$,�}`QX�+��W@� �4,��q�w��1 �q�cb!��H�B\��/��Ÿ��_~��2�W|F9�",w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �Q��0;�s� �e�p�c0��`��q�b0�($r�8H1�`��`�@��\0S ��L.�)��'��v@~�Ƒ��@;��P�` �|�q�b0�(�r�8X1�`$v�`�@�ڿ\0W ��b��@;��r�8`1�`d�A.G,�~5��` ��`�@��c��v@��r�c0��Y@t ��x@ȑ��@;�qd9�1�`T��+G-�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0] �� Ʊ��@;��\�8x1�`TƂ+G/��Yp��` �Ph�`�@�zq��v@�.��q�c0�h�W0q �� 1��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]�K�H��|�@�Ʌ�0H�Ho�x�@��"�3��C��f�Z `�d�W�D�Y��!Rk3�|�0D�mfZ�H��X��i��U"�8�� C�;g�U `�d�̵*@<�z��!R�3�0Dy��H0�@�@� �)>"�=I�H�a��!�3�|0D�}��H��<�@�I�i>"�`#�C$l�� `��M4�@�ad��!+6��0D��F��H��D�@��V"�d�� C$�lZ `�T��C+�貑h��!�i6�0D��&��H��D�@�I?�V"�h�� C�m$[ `�t�Md+��ah��!�6�0D��H��<�@��]�V"�l#� C$�m"Z `�4�MD+�趙p��!��6��0D��f��H��X�@��W"�p�� C�:n0\ `��ͅ+�ȹ�p��!�>7�0Dr�� H<�8�@�I��W"�v�� C$�n.^ `��.�D��!R�7�0DZ��H��`�@�)�\"�{s�C$�o.p `��$.�侱��!�7��0D��H��L�@�I� \"��c�C�9t,p `��.��!R>:�0D�I��Hg�,�@�I.��\"��C$�ts `�4�Β.D�|�Y �H��,�@��>�]"��C$u"y `�T�NT/��y��!R�:Ͼ0D�U��H��,��i^�%]�"��@�R�Y��!�:Q�0D*['"�Hv�H'@� uIe"��C$�u � `��53�`�Yj��!�;��0D2c'��Hw�D=@� � l"�ÐC$svr `�$�c.D��a��!�J;L�0D�i��"��ÐC$�v"v `��N�.�`ۑ��!�x;R�0D�o��H$�@�@��Ɲ�_"��ØC�:w�t@`��0Dju�!�H��H!@�i��d"��3�C$�w&� `�4�N�2��މ`��!R�;��0D�}��H��07@��Ig"��#�C$|&� `��C.��a��!�(>��0DzŇ1"�H��0�@�� `�4�C.��!L>ѻ0D �G��Hs�H�@��4_"��C��|* `��c.H�L�a� �!��>@��G�\"��#�C$>}$� `��t2��T��!�>Q�0D*�'��H6�\3@��m�f"��C��}$� `�t��3��!�>�0DB�!�H4�0�@� ��\�"��äC$<~��!)?�0D��'b�H��D�@�I�I^"��#�C$�~ | `�d��/��!�j?�� 0D��I$�H��0C��~r `��ߏ2�Z��H��!ҏ?��0D��gR�H��D-@��֟f"��s�C$�� `��s3�l��t��!�?R�0D��g�H��0�@�)��\"��ØC��s `��.D�Rb �H51�@�)��]"�!�C�Q�$y `�T��T/��!�=Bо0D:I��H< 1�� -!&]�"�%�@�@b��!kBR�0D�MH"�H� M'@�iD�Ie"�(�C�3�"� `��53�^bj��!ҮB��0D:WH��H� I=@�ie� l"�,ĐC$��r `�d�c.D�Db��!��BL�0D�]��"�/ĐC$��"v `�4�P�.�h��!�CR�0D�c��H�A�@� ��_"�3ĘC�|��t@`�T�0D�i�!�H= I!@��!�d"�64�C$چ&� `�t�P�2��`��!R�C��0Ds��Hl17@� �!Ig"�:$�C$^�&� `��#6��b��!��C�00D�y�!&�H?1f��i�!m"�=Ĉ C�ч$� `�T��6��!��C�0Dj�(��H>Et@�� #�n"�bĐ C�X��`�ԋ�6��1b��!R=F�0DB�h��H;Q|@��-#�o"�f$�C�ڌ$� `�t�F:��3bL��!ҊF�0D��hR�HjM�@��O#�u"�jĈ C�^�� `��C6L��5b�� !R�F��00DjوA�H91b@��l#Im"�m$� C�ԍ&� `�Ļ�7��7��!G�0D��Ȳ�Hg1d�i�#�l�"�qĠ C�U�� `�t�ф7��9��!R?G�0D��HeI�@��#It"�v��C�� `��:��;�T��!�G�0D��r�Hs1b@��#Fl"�zĐ C�[��``�4�c6L��=b��!ҾG��0D:�HR�H9Il@�i��m"�}4� C�؏"� `�$�QD7��?��!R J�@0D�A�1(�Ha(1h@�� %Fl"�4� C$V�&� `��7��Q��!�IJ�0D�II�Hk)a�@�i4%�t"��ĠC�ݔ&� `��:�BT�\��!R�J��0D S��Hm*1d��)O%�l�"��Ę C�X�� `�ԫ#6��U��!ҾJ�0DjYir�H`+Mp@�in�Hn"�� C�ڕ,� `��C6P��Wb� �!�K�0Dza��H;,Mx@�ɍ�Io"��D� C�Z�(� `��Β:��Y�D��!RKK�0D*k�1�Hq-1�@��Iu"��4�C��(� `�tߒ5�Z\��!�K�|0D�s�$6��$2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B� b D�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align�+0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1��..6�//6�0/06�`/16��/26��/36��/46� 296�P2:6��2;6��2<6��2=6�3>6�@3?6��3b6��3c6�6d6��.l6�/m6�0/n6�`/o6��/p6��/q6��/r6� 2w6�P2x6��2y6��2z6��2{6�3|6�@3}6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.&7�/'7�0/(7�`/)7��/*7��/+7��/,7� 217�P227��237��247��257�367�@377��3:7��3;7�6<7��.d7�/e7�0/f7�`/g7��/h7��/i7��/j7� 2o7�P2p7��2q7��2r7��2s7�3t7�@3u7��3x7��3y7�6z7��6�7��a0��6�7��a0��6�7�7�7��a0�07�7�P7�7��b0�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6 � �M#s#��B݌`0�p7$� �M #{��B߬`0��7,� �N#��B�`0��:2� ��#�C��`0��6=� ��#q��܀a0�P7a� ܍#ys��ޠa0��7i� ��#��a0�P:m� �#��a0� 3D �0�0�`�3Ip�0(�1ð�C03N��0<3��D3S�0PA4�P��D3W��0`A4Ð��D3[�0pA4��D3_3a��`0�0�0�A3e��`0�p�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�D��0�D��0�D�0�D�0�D �0�Dp�0�Dp�0�D��0�D��0�D�0�D�0�D��0�D�0�D�0�D�30��L11�0M�0�D@36��N71Ð�K�0�D3;��O41��M�0�Dp3?��X41��M�0�Ep3c��Y11�PAL�0�E@3g��Z71Ð�M�0�E�3k��[11��M�0�E@3o��\71�M�0�Ep3s��]/1�PAL�0�E3w��^41Ð�M�0�Ep3{��_11��AL�0�E@3��h71��M�0�F@3��i71�P�K�0�F3��j41ÐM�0�Fp3��k/1��AL�0�F3��l41��M�0�Fp3��m71�P�M��h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n��H`��q��kK�s�+��K{s%��\��FQ ۸ �� (�8�H�X�h�x�� )9IYiy��0 12)394I5Y6i7y8�9�:�;�<�=�>�?�P QR)S9TIUYViWyX�Y�Z�[�\�]�^�_�p qr)s9tIuYviwyx�y�z�{�|�}�~�� )�9�I�Y�i�y�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d��\Ǝ��4Ma A,$�� d��\Ǝ��4Ma A,T�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,D�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,t�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� d��\Ǝ��4Ma A,d�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,t�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,4�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� 6C�A8 ��\��`֎��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,$�� 6C�A8 ��\��`֎��4M��4Ma A,T�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,T�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a �� G�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|ğ\p��A|ԟ��A|ܟ\p�@�@T� WC��5֏`NSٽ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D&�@̀[<B!8Me�<B!8M�>B!Q}[=B!8M��;!8Me��4U][>�!8MU��`��ܷ&��Txo� ��4U��:B�:B;B =!�n� �,na�!˂�f�`NS��#@3Դ�#��T�m��4��?!�a �� { �� G{B�@��Ǝ֎`��#�,>Ra �� z A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|ğ\p�@@��A|ԟ�@��A|ܟ\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�%&Nd� ��4��`�!�r�&�@̀[<B!8Me[=B!8M��<B!8M��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}?!�a �� G� �� ć�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5֏`NSٽ�#�,��`NSٶ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D�PNS�V�`NS��#��T��#@3��PHGT��@NSY=B!8M��=�!8MUז�`NSո&��T�m� ��4U��:B�:B;B =!�n��,�}�`�!˂�F�`��#@3Դ�#��T6a �� ć �� B�@��Ǝ֎`��#�,>Ra �� G A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�&��T�n� �,˭�<4n��4�m��4Un��4o� �>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}?!�a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|ȟ\p��A|؟��A|��\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a � ��A|�_ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|�\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m� �,��<B!8M�[=B!8M�>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�^ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�_ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me[>B!Q}=B!8M�<!8Me��4U]�>�!8MU�f�PNS�`NSἩ#��#��#��#�햘`��9&��Tym� ��4U�[?��I&��Tyo� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a � ��A|�^ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me=B!8M��=B!8M�[>B!Q}<!8Me��4U]�>�!8MU��`NS�-&��T8o�`�A`��`� ��4U�[=!�n� �,�[d�!8M��֏0�`� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a �� p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� { �p1��Q�D8�0�c�`�A`��#��T��#��T֎`��a �� G� �p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� ć �p1��Q�D8�0�c�`�A`��4�n��#��T֎`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��~Bp�p r2b` �Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��~Bp�p q1b` �Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��@��a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� H� A�@��Ǝ@��a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a �� H� A�@��Ǝ@��a �� ȿ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� Ȼ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� ��@��a ��A� 䇹��@��a �� H� A�@��Ǝ@��a �� ȸ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� Ȭ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ȩ A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� Ȩ A�@��Ǝ@��a �� B�A8��Ǝ0@�c��,a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a �� ț A�@��Ǝ@��a �� Ј� B�A8��Ǝ֎@��a �� Ț A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ؈� B��@��@H�;!<Sa �� B��@��@H�;!<Sa ��A� ��@��a ��A� ��@��a �� H� A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a �� ȋ A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� A�@��Ǝ@NSY;!�a � �� B�@��Ǝ@NSY;!�a �� G A�@��Ǝ@��a �� G A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� I~ A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� { A�@��Ǝ@��a � �� {B�@��Ǝ@NSY;!�a �� { A�@��Ǝ@NSY;!�a �� z A�@��Ǝ@��a �� z A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� y A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Ț A�@��Ǝ@��a �� o A�@��Ǝ@��a � �� HoB�@��Ǝ@NSY;!�a �� Ho A�@��Ǝ@NSY;!�a �� n A�@��0T�b�0!IƎ@��a �� m A�@��0T�b�0!IƎ@��a ��A#�` &�5��0T�b�0!IƎ04�a��,a ��A#�` &�5��0T�b�0!IƎ04�a��,a �� l A�@��0T�b�0!IƎ@��a �� l A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ] A�@��0T�b�0!IƎ@��a �� ,u5A�@Ǝ0T�b�0!I��p�a��,a �� ,s5A�@Ǝ0T�b�0!I��p�a��,a �� \ A�@��0T�b�0!IƎ@��a �� [ A�@��0T�b�0!IƎ@��a �� I[ B�A8Ǝ0T�b�0!I��p�a��,a �ȈA� ��Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a �ȈA� � Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a ;J,@ �� a��pC�`0�0�� [%�1K�7�2F0bp p�V��#�`'m�,�,�1PAD!��@y(�p�$��A86�`��Խ�#�<B!0OM>�!8̀Ӗ�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#f��p�Q��,Ca#�`�#g�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3v�`N3ؼ�#��@̳>�!8�`�F�`��ص�#�:B�:BX=�!�b�6�P�S��@��a 9J,@ �� Q�$� E��,�@#�`�#a��p�Q��,Ca#�`�#b�$�ǈ�� X-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��>�!8�`��@��a 9J,@ �� Q�$� E��,�@#�`�#<��p�Q��,Ca#�`�#=�$�ǈ�� HO-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��<!�b��4��a 5J,@� ��,C #�`�4E�$� ��,�@#�`�5e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�/E�$� ��,�@#�`�0e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'*E�$� ��,�@#�`�'+e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'%E�$� ��,�@#�`�'&e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP |2R�,�1bP |2R�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a � �X�@��@0�@;!8Ma � �T�@��@0�@;!8Ma � �`�@��@0�@;!8Ma �� I< A�@��0T�b�0!IƎ@��a �� ; A�@��0T�b�0!IƎ@��a �� I; A�@��0T�b�0!IƎ@��a �� : A�@��0T�b�0!IƎ@��a �� I: A�@��0T�b�0!IƎ@��a �� 9 A�@��0T�b�0!IƎ@��a �� I9 A�@��0T�b�0!IƎ@��a �� 8 A�@��0T�b�0!IƎ@��a �� I8 A�@��0T�b�0!IƎ@��a �� / A�@��0T�b�0!IƎ@��a �� I/ A�@��0T�b�0!IƎ@��a �� . A�@��0T�b�0!IƎ@��a �� I. A�@��0T�b�0!IƎ@��a �� - A�@��0T�b�0!IƎ@��a �� I- A�@��0T�b�0!IƎ@��a �� , A�@��0T�b�0!IƎ@��a �� I, A�@��0T�b�0!IƎ@��a �� + A�@��0T�b�0!IƎ@��a �� I+ A�@��0T�b�0!IƎ@��a �� * A�@��0T�b�0!IƎ@��a �� J* A�@��0T�b�0!IƎ@��a �� ) A�@��0T�b�0!IƎ@��a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a �� I( A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� ,*4A�@Ǝ0T�b�0!I��p�a��,a �� ,(4A�@Ǝ0T�b�0!I��p�a��,a �� ,&4A�@Ǝ0T�b�0!I��p�a��,a �� ,$4A�@Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �ȈA� �I��@NS�:�D�`��,a �ȈA� ��@NS�:�D�`��,a �ȈA� �I�� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a �ȈA� �� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a ��A� � ��0<Sa ��A� ��0<Sa ��A� � ��0<Sa ��A� ��0<Sa ��d�@��<a ��d�@��<a ��d�@��<a ��d�@��<a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a �� ̇�B�A8Ǝ`8E!��# )��@�Da ��A� �� a �� J��0�d�2a ��A� ��@0��a �pCPa0bp ��6�� Ap�A8��@<1�:��a A,@ � �� B0��@<1�:��a �pCPa0bp ��6C� Ap�A8��@<1�:��a A,@� �� B0��@<1�:��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| dƎPX�0�PHT�#��pH�P a�CD&�@8��X<B!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ ƎPX�0��PHT�#��0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��#�L��#a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��@,`��#ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��4� H��#N�#�PH��#a &A,@�M5��<��9��9��A��6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa #A,@�Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa %A,@�M5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��#�L��#a &A,@�M5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��m� 1!Q�T[>�!Q��V�PHT�#v�@8��>�!��f�PH��#6��L��0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�PHT�#v�`HT�#��#(��D�D-Sm��t,>b[=B!��<BHLDH�2�:�P��#�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#��P8��ܦ�`H��#��#(��D�D-Sm��#�Hm��t,>b� 1!Q�d�CD��P�2a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��#ұ��<�HLDH�2��# �H�#a UA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#F��L��#ұ��m��#N�#��#�Hm��t,>b� 1!Q�d�CD��,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#��HT�#��=�!��#,��D�D-��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m�� %��HT�#>R=�HLDH�2զ��H��#>b[=�!Q��#ұ��=��D>�#6��L��0T@a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@dƎPX�0v��HT�#>R>"��m� 1!Q�T=�!Q��#ұ��<BHLDH�2�:�P��#�La WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�:�P��#�L��#N�#>b��t,>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�=�!��#�:� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CDV��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0v��HT�#��#$ұ��#4��D�D-��#�:B1�=B"Q��>�"��>�#6��L��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��# �H�#f�pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��@,`��D>�#>b��#>b��t,>b��=�HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��4� H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ;�!EB 4��#<� Q��# )a A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#ƎpH�P a��#ұ��#�:��a A,@�&!P�܈A� �J� �a� �A8 ֎PX�0�0H�b�CDƎP�2��@0��<� Qa A,@�&!P�܈A� �J� �a� �A8֎PX�0�PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ�:�P��#�L��# )�0HTa A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#&�@8��X<B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��֎PX�06� 8��#ұ�:�P��# �& )��@0��X<� Qa A,@�O6!P\Ď$��B$H0�� ֎PX�0�PHT�#6�@8��<B!��:�P��# )Ǝ�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d֎PX�0&� 8��#ұ;�!EB 4��#�,�m�CD��@0��<� Q��#�,a A,@�&!P�܈A� ��# �`�+.��!�&��A� �� ֎PX�0V�`��Է�#ұ�:�P��#�L��# )�0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d֎PX�0&� 8��#ұ�=�!�R߶�0T@d��@,`�H��#�<!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8֎PX�06� 8��#ұ�=�!�R߶�0T@d�H��# �& )�0HT`��,a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�>�!Q��v�@8��<BHLDH�2�:�Pa MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a FA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�PHT�#��`H��#��#(��D�D-Sm��t,>b�H��#$��D�D-��#�Hm�CDƎP�2a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#�Hm� 1!Q�d��#;B1�a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a JA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�PHT�#֏P8��Ə`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#N�#F��L��0T@a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��;�!EB 4��# ��Զ��H��#>b?��D>�#��#$��D�D-��#�=�!��a GA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ~�E�@֎PX�0F��L��#��:��b��D>�#��# ұ��6��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��?��D>�#��# ��Զ��H��#>b�<BHLDH�2�:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<wsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@֎PX�0V��L��#��:��b��4� H��#N�#>b��ܶ��HT�#>R?"��m� 1!Q�d�CDv�pH��#>a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CD��@0��X=B"Q��?B�D>r��>�"��>�#F��L��a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0��HT�#��#$ұ��#4��D�D-��#8��D�D-S�#>b�CDƎP�2��@0��X=B"Q��>�"��>�#a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a PA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#֏�8��>�#��0T@d�H��# �& )f��HT�#>�#Ə�H��#��#8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q�ě:��L��#��T5n��4U͛;BHLDH�2�<!8Me��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�Y<�!8MU��@NS�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#��T��L��#�<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2�&��4MF��4M��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��_�O�H#�d��4UM�<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!v��4M��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �q }2"�\�3�U<�^�3��>�`�3�I�b�48FM�d�4XVX�f�5x�Z�h�5��]�j�5�Fi�l�6�Vl�n�6��n�p�6�y�r�78G}�t�7XW��v�8x��x�8��z�8�G��|�9�W��~�9�G��:P��:8��:X��;x@��;�P��;�@Ƀ��<��̃��=�ۃ��>8pꃔ�?X`��?x��@��A��A�@(��B�@9��D�I��D8�X��EX�_��Fxo��G��~��H��I��J�𫄰�J��L80؄��Mh0߄��M��脼�N�0��N(��Nh0��N��N�0��O��OH0��O��O�@��O��PH` ��P��P��P ��PH��Q� ��Q��Q(0��QH(��Q�(@��R�(�(��R)P*��RH)�+��R�)P-��R�)�.��R*`8��SH*�9��S�*�;��S�*@=��S+�>��SH+@H��T�+�I��T�+@K��T,�L��TH,�N��T�,X��U�,�Y��U-0[��UH-�\��U�-`^��U�-�_��V.pi��VH.�j��V�.pl��V�.�m��V/�o��W(/py��Wh/�}��X�/p��X�/��X�/��X�/P��Y(80��YH8К��Y�8P��Z�8Ы��Z�8p��Z�8��Z9��[89��[x9p��[�9��[�9pȅ��\:Pʅ��\X:0̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<p��_�<��_�=P��`8>0��a�>)��bx?08��c�?�9��c�?`;��cH@=��cHH ?��d�HI��d�H�J��dI�L��dHI�N��d�I�X��e�I`Z��eJ@\��eHJ ^��exJ h��f�J@j��f�J@l��fK n��fHKx��g�K�y��g�K�{��gL�}��gHL��h�L`��h�L@��hM ��hHM��h�M��i�M��iN��iHN��i�N��j�Nઆ��jOଆ��jHO��j�O��k�O��kX`��kHX@��k�X Ȇ��l�Xʆ��lY�ˆ��lHY�͆��l�Y�φ��m�Y�ن��mZPۆ��mHZ�݆��m�Z�߆��n�Z醮�n�Z@ꆰ�n[`놲�n8[�솴�nX[�톶�n�[`�o�[��o\`��o(\��oH\��ph\��p�\��q�\��r�\�*��r�\�/��s]p<��t(]�O��uH]�i��vh]�|��x�]��y�]��{�]��|�]ه��}^��(^��ɀH^`��x^��^@)��^�,��^�8��_�<��8_�I��X_0[��x_Pl��_~��_��ʉ�_0��_��h��8hp��Xh��׏xh� ��Đ�h� ��hp��h��h�)��ip9��ړ8i�H��͔Xi�O��xi@[��Ε�i�^��i�n��i�}��i��j��8jP��Xj��xj��j`��j��j��j`��k��8k��ȟXk0��xk� ��ˠ�k��ȡ�k�(��k�/��k�<��l@I��פ8lpX��Xl�[��٥xlP_��l�j��Ŧ�l�o��ŧ�lP��l��m��8m ��Xmp��xmp��m��m��ݭ�m��m0��n`��ѯ8n��Xn@ ��xn� ��np��n@*��np9��ڳ�n�H��oL��۴8o`O��Xo`_��xo�n��o�}��op��¸�o ��oP��x��8x��Xx��xx��¼�x`��x��x��x��ǿyP��8y��Xy��xy0.��yP;��ypH��y X��y�[��z�_��8z�k��Xz�x��xz0��z��z��zp��z ��{��8{P��X{��x{ ��{@��{� ��{� ��{��|��8|0+��X|p:��x| I��|�O��|�[��|�_��|`o��}~��8}��X}��x}��}0��}0��}P��}P��~p��8~��X~`��x~��~ ��~ ��~��~P)��8��8�<��X�I��xpX��p\��h��l��y��8��X�P��x�`��`��؈��@�� 8��X��x��؉��0,��:��8��I��X��M��x��Y�� i��o��؊�~��`��@��8��X��]��.threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_60_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU�_~��DU!�!ȪR@X�)H�R��/@\�ȫR�@`��4@d�� ,��w�EiQE(c-b��(Bn�[�G��"E��]TYRK�"�(�"��E�A��/B�hF�E�c0"�)B ��a�HI��L�� c��RR#Ȩ�"T#MV�_�ʊ@�Pt S�a2#�(�"$�uF��2\р$Z��"�W�u�V�+�P+�k�A�� pE\~�ke��.E�2\q�%$[��+��W�% �V�+C�p+��A�� !��pEb��oe��L�2\�)$\��"�W$�W�+�"��+�� Cʕ� Β!��p�g�sS�T0�@ �1�@F�"Xg]W��+�u��vE�n-��E��-�w�Q��q�w�Q��"�`�]�^�Ӌ�3��"X�W�� a4_�K��+�u��}E��1ʯ�IF��_��_f�!~��_� g��!�~F�� q�Z@� �H�B\�w�Ÿ��;��r�W�a1��$,�}`QX�+��W@� �4,��q�w��1 �q�cb!��H�B\��/��Ÿ��_~��2�W|F9�",w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �Q��0;�s� �e�p�c0��`��q�b0�($r�8H1�`��`�@��\0S ��L.�)��'��v@~�Ƒ��@;��P�` �|�q�b0�(�r�8X1�`$v�`�@�ڿ\0W ��b��@;��r�8`1�`d�A.G,�~5��` ��`�@��c��v@��r�c0��Y@t ��x@ȑ��@;�qd9�1�`T��+G-�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0] �� Ʊ��@;��\�8x1�`TƂ+G/��Yp��` �Ph�`�@�zq��v@�.��q�c0�h�W0q �� 1��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]�K�H��|�@�Ʌ�0H�Ho�x�@��"�3��C��f�Z `�d�W�D�Y��!Rk3�|�0D�mfZ�H��X��i��U"�8�� C�;g�U `�d�̵*@<�z��!R�3�0Dy��H0�@�@� �)>"�=I�H�a��!�3�|0D�}��H��<�@�I�i>"�`#�C$l�� `��M4�@�ad��!+6��0D��F��H��D�@��V"�d�� C$�lZ `�T��C+�貑h��!�i6�0D��&��H��D�@�I?�V"�h�� C�m$[ `�t�Md+��ah��!�6�0D��H��<�@��]�V"�l#� C$�m"Z `�4�MD+�趙p��!��6��0D��f��H��X�@��W"�p�� C�:n0\ `��ͅ+�ȹ�p��!�>7�0Dr�� H<�8�@�I��W"�v�� C$�n.^ `��.�D��!R�7�0DZ��H��`�@�)�\"�{s�C$�o.p `��$.�侱��!�7��0D��H��L�@�I� \"��c�C�9t,p `��.��!R>:�0D�I��Hg�,�@�I.��\"��C$�ts `�4�Β.D�|�Y �H��,�@��>�]"��C$u"y `�T�NT/��y��!R�:Ͼ0D�U��H��,��i^�%]�"��@�R�Y��!�:Q�0D*['"�Hv�H'@� uIe"��C$�u � `��53�`�Yj��!�;��0D2c'��Hw�D=@� � l"�ÐC$svr `�$�c.D��a��!�J;L�0D�i��"��ÐC$�v"v `��N�.�`ۑ��!�x;R�0D�o��H$�@�@��Ɲ�_"��ØC�:w�t@`��0Dju�!�H��H!@�i��d"��3�C$�w&� `�4�N�2��މ`��!R�;��0D�}��H��07@��Ig"��#�C$|&� `��C.��a��!�(>��0DzŇ1"�H��0�@�� `�4�C.��!L>ѻ0D �G��Hs�H�@��4_"��C��|* `��c.H�L�a� �!��>@��G�\"��#�C$>}$� `��t2��T��!�>Q�0D*�'��H6�\3@��m�f"��C��}$� `�t��3��!�>�0DB�!�H4�0�@� ��\�"��äC$<~��!)?�0D��'b�H��D�@�I�I^"��#�C$�~ | `�d��/��!�j?�� 0D��I$�H��0C��~r `��ߏ2�Z��H��!ҏ?��0D��gR�H��D-@��֟f"��s�C$�� `��s3�l��t��!�?R�0D��g�H��0�@�)��\"��ØC��s `��.D�Rb �H51�@�)��]"�!�C�Q�$y `�T��T/��!�=Bо0D:I��H< 1�� -!&]�"�%�@�@b��!kBR�0D�MH"�H� M'@�iD�Ie"�(�C�3�"� `��53�^bj��!ҮB��0D:WH��H� I=@�ie� l"�,ĐC$��r `�d�c.D�Db��!��BL�0D�]��"�/ĐC$��"v `�4�P�.�h��!�CR�0D�c��H�A�@� ��_"�3ĘC�|��t@`�T�0D�i�!�H= I!@��!�d"�64�C$چ&� `�t�P�2��`��!R�C��0Ds��Hl17@� �!Ig"�:$�C$^�&� `��#6��b��!��C�00D�y�!&�H?1f��i�!m"�=Ĉ C�ч$� `�T��6��!��C�0Dj�(��H>Et@�� #�n"�bĐ C�X��`�ԋ�6��1b��!R=F�0DB�h��H;Q|@��-#�o"�f$�C�ڌ$� `�t�F:��3bL��!ҊF�0D��hR�HjM�@��O#�u"�jĈ C�^�� `��C6L��5b�� !R�F��00DjوA�H91b@��l#Im"�m$� C�ԍ&� `�Ļ�7��7��!G�0D��Ȳ�Hg1d�i�#�l�"�qĠ C�U�� `�t�ф7��9��!R?G�0D��HeI�@��#It"�v��C�� `��:��;�T��!�G�0D��r�Hs1b@��#Fl"�zĐ C�[��``�4�c6L��=b��!ҾG��0D:�HR�H9Il@�i��m"�}4� C�؏"� `�$�QD7��?��!R J�@0D�A�1(�Ha(1h@�� %Fl"�4� C$V�&� `��7��Q��!�IJ�0D�II�Hk)a�@�i4%�t"��ĠC�ݔ&� `��:�BT�\��!R�J��0D S��Hm*1d��)O%�l�"��Ę C�X�� `�ԫ#6��U��!ҾJ�0DjYir�H`+Mp@�in�Hn"�� C�ڕ,� `��C6P��Wb� �!�K�0Dza��H;,Mx@�ɍ�Io"��D� C�Z�(� `��Β:��Y�D��!RKK�0D*k�1�Hq-1�@��Iu"��4�C��(� `�tߒ5�Z\��!�K�|0D�s�$6��$2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B� b D�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align�+0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1��..6�//6�0/06�`/16��/26��/36��/46� 296�P2:6��2;6��2<6��2=6�3>6�@3?6��3b6��3c6�6d6��.l6�/m6�0/n6�`/o6��/p6��/q6��/r6� 2w6�P2x6��2y6��2z6��2{6�3|6�@3}6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.&7�/'7�0/(7�`/)7��/*7��/+7��/,7� 217�P227��237��247��257�367�@377��3:7��3;7�6<7��.d7�/e7�0/f7�`/g7��/h7��/i7��/j7� 2o7�P2p7��2q7��2r7��2s7�3t7�@3u7��3x7��3y7�6z7��6�7��a0��6�7��a0��6�7�7�7��a0�07�7�P7�7��b0�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6 � �M#s#��B݌`0�p7$� �M #{��B߬`0��7,� �N#��B�`0��:2� ��#�C��`0��6=� ��#q��܀a0�P7a� ܍#ys��ޠa0��7i� ��#��a0�P:m� �#��a0� 3D �0�0�`�3Ip�0(�1ð�C03N��0<3��D3S�0PA4�P��D3W��0`A4Ð��D3[�0pA4��D3_3a��`0�0�0�A3e��`0�p�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�D��0�D��0�D�0�D�0�D �0�Dp�0�Dp�0�D��0�D��0�D�0�D�0�D��0�D�0�D�0�D�30��L11�0M�0�D@36��N71Ð�K�0�D3;��O41��M�0�Dp3?��X41��M�0�Ep3c��Y11�PAL�0�E@3g��Z71Ð�M�0�E�3k��[11��M�0�E@3o��\71�M�0�Ep3s��]/1�PAL�0�E3w��^41Ð�M�0�Ep3{��_11��AL�0�E@3��h71��M�0�F@3��i71�P�K�0�F3��j41ÐM�0�Fp3��k/1��AL�0�F3��l41��M�0�Fp3��m71�P�M��h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n��H`��q��kK�s�+��K{s%��\��FQ ۸ �� (�8�H�X�h�x�� )9IYiy��0 12)394I5Y6i7y8�9�:�;�<�=�>�?�P QR)S9TIUYViWyX�Y�Z�[�\�]�^�_�p qr)s9tIuYviwyx�y�z�{�|�}�~�� )�9�I�Y�i�y�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d��\Ǝ��4Ma A,$�� d��\Ǝ��4Ma A,T�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,D�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,t�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� d��\Ǝ��4Ma A,d�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,t�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,4�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� 6C�A8 ��\��`֎��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,$�� 6C�A8 ��\��`֎��4M��4Ma A,T�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,T�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a �� G�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|ğ\p��A|ԟ��A|ܟ\p�@�@T� WC��5֏`NSٽ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D&�@̀[<B!8Me�<B!8M�>B!Q}[=B!8M��;!8Me��4U][>�!8MU��`��ܷ&��Txo� ��4U��:B�:B;B =!�n� �,na�!˂�f�`NS��#@3Դ�#��T�m��4��?!�a �� { �� G{B�@��Ǝ֎`��#�,>Ra �� z A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|ğ\p�@@��A|ԟ�@��A|ܟ\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�%&Nd� ��4��`�!�r�&�@̀[<B!8Me[=B!8M��<B!8M��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}?!�a �� G� �� ć�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5֏`NSٽ�#�,��`NSٶ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D�PNS�V�`NS��#��T��#@3��PHGT��@NSY=B!8M��=�!8MUז�`NSո&��T�m� ��4U��:B�:B;B =!�n��,�}�`�!˂�F�`��#@3Դ�#��T6a �� ć �� B�@��Ǝ֎`��#�,>Ra �� G A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�&��T�n� �,˭�<4n��4�m��4Un��4o� �>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}?!�a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|ȟ\p��A|؟��A|��\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a � ��A|�_ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|�\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m� �,��<B!8M�[=B!8M�>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�^ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�_ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me[>B!Q}=B!8M�<!8Me��4U]�>�!8MU�f�PNS�`NSἩ#��#��#��#�햘`��9&��Tym� ��4U�[?��I&��Tyo� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a � ��A|�^ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me=B!8M��=B!8M�[>B!Q}<!8Me��4U]�>�!8MU��`NS�-&��T8o�`�A`��`� ��4U�[=!�n� �,�[d�!8M��֏0�`� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a �� p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� { �p1��Q�D8�0�c�`�A`��#��T��#��T֎`��a �� G� �p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� ć �p1��Q�D8�0�c�`�A`��4�n��#��T֎`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��~Bp�p r2b` �Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��~Bp�p q1b` �Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��@��a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� H� A�@��Ǝ@��a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a �� H� A�@��Ǝ@��a �� ȿ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� Ȼ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� ��@��a ��A� 䇹��@��a �� H� A�@��Ǝ@��a �� ȸ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� Ȭ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ȩ A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� Ȩ A�@��Ǝ@��a �� B�A8��Ǝ0@�c��,a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a �� ț A�@��Ǝ@��a �� Ј� B�A8��Ǝ֎@��a �� Ț A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ؈� B��@��@H�;!<Sa �� B��@��@H�;!<Sa ��A� ��@��a ��A� ��@��a �� H� A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a �� ȋ A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� A�@��Ǝ@NSY;!�a � �� B�@��Ǝ@NSY;!�a �� G A�@��Ǝ@��a �� G A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� I~ A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� { A�@��Ǝ@��a � �� {B�@��Ǝ@NSY;!�a �� { A�@��Ǝ@NSY;!�a �� z A�@��Ǝ@��a �� z A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� y A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Ț A�@��Ǝ@��a �� o A�@��Ǝ@��a � �� HoB�@��Ǝ@NSY;!�a �� Ho A�@��Ǝ@NSY;!�a �� n A�@��0T�b�0!IƎ@��a �� m A�@��0T�b�0!IƎ@��a ��A#�` &�5��0T�b�0!IƎ04�a��,a ��A#�` &�5��0T�b�0!IƎ04�a��,a �� l A�@��0T�b�0!IƎ@��a �� l A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ] A�@��0T�b�0!IƎ@��a �� ,u5A�@Ǝ0T�b�0!I��p�a��,a �� ,s5A�@Ǝ0T�b�0!I��p�a��,a �� \ A�@��0T�b�0!IƎ@��a �� [ A�@��0T�b�0!IƎ@��a �� I[ B�A8Ǝ0T�b�0!I��p�a��,a �ȈA� ��Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a �ȈA� � Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a ;J,@ �� a��pC�`0�0�� [%�1K�7�2F0bp p�V��#�`'m�,�,�1PAD!��@y(�p�$��A86�`��Խ�#�<B!0OM>�!8̀Ӗ�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#f��p�Q��,Ca#�`�#g�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3v�`N3ؼ�#��@̳>�!8�`�F�`��ص�#�:B�:BX=�!�b�6�P�S��@��a 9J,@ �� Q�$� E��,�@#�`�#a��p�Q��,Ca#�`�#b�$�ǈ�� X-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��>�!8�`��@��a 9J,@ �� Q�$� E��,�@#�`�#<��p�Q��,Ca#�`�#=�$�ǈ�� HO-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��<!�b��4��a 5J,@� ��,C #�`�4E�$� ��,�@#�`�5e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�/E�$� ��,�@#�`�0e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'*E�$� ��,�@#�`�'+e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'%E�$� ��,�@#�`�'&e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP |2R�,�1bP |2R�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a � �X�@��@0�@;!8Ma � �T�@��@0�@;!8Ma � �`�@��@0�@;!8Ma �� I< A�@��0T�b�0!IƎ@��a �� ; A�@��0T�b�0!IƎ@��a �� I; A�@��0T�b�0!IƎ@��a �� : A�@��0T�b�0!IƎ@��a �� I: A�@��0T�b�0!IƎ@��a �� 9 A�@��0T�b�0!IƎ@��a �� I9 A�@��0T�b�0!IƎ@��a �� 8 A�@��0T�b�0!IƎ@��a �� I8 A�@��0T�b�0!IƎ@��a �� / A�@��0T�b�0!IƎ@��a �� I/ A�@��0T�b�0!IƎ@��a �� . A�@��0T�b�0!IƎ@��a �� I. A�@��0T�b�0!IƎ@��a �� - A�@��0T�b�0!IƎ@��a �� I- A�@��0T�b�0!IƎ@��a �� , A�@��0T�b�0!IƎ@��a �� I, A�@��0T�b�0!IƎ@��a �� + A�@��0T�b�0!IƎ@��a �� I+ A�@��0T�b�0!IƎ@��a �� * A�@��0T�b�0!IƎ@��a �� J* A�@��0T�b�0!IƎ@��a �� ) A�@��0T�b�0!IƎ@��a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a �� I( A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� ,*4A�@Ǝ0T�b�0!I��p�a��,a �� ,(4A�@Ǝ0T�b�0!I��p�a��,a �� ,&4A�@Ǝ0T�b�0!I��p�a��,a �� ,$4A�@Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �ȈA� �I��@NS�:�D�`��,a �ȈA� ��@NS�:�D�`��,a �ȈA� �I�� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a �ȈA� �� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a ��A� � ��0<Sa ��A� ��0<Sa ��A� � ��0<Sa ��A� ��0<Sa ��d�@��<a ��d�@��<a ��d�@��<a ��d�@��<a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a �� ̇�B�A8Ǝ`8E!��# )��@�Da ��A� �� a �� J��0�d�2a ��A� ��@0��a �pCPa0bp ��6�� Ap�A8��@<1�:��a A,@ � �� B0��@<1�:��a �pCPa0bp ��6C� Ap�A8��@<1�:��a A,@� �� B0��@<1�:��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| dƎPX�0�PHT�#��pH�P a�CD&�@8��X<B!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ ƎPX�0��PHT�#��0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��#�L��#a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��@,`��#ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��4� H��#N�#�PH��#a &A,@�M5��<��9��9��A��6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa #A,@�Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa %A,@�M5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��#�L��#a &A,@�M5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��m� 1!Q�T[>�!Q��V�PHT�#v�@8��>�!��f�PH��#6��L��0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�PHT�#v�`HT�#��#(��D�D-Sm��t,>b[=B!��<BHLDH�2�:�P��#�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#��P8��ܦ�`H��#��#(��D�D-Sm��#�Hm��t,>b� 1!Q�d�CD��P�2a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��#ұ��<�HLDH�2��# �H�#a UA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#F��L��#ұ��m��#N�#��#�Hm��t,>b� 1!Q�d�CD��,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#��HT�#��=�!��#,��D�D-��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m�� %��HT�#>R=�HLDH�2զ��H��#>b[=�!Q��#ұ��=��D>�#6��L��0T@a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@dƎPX�0v��HT�#>R>"��m� 1!Q�T=�!Q��#ұ��<BHLDH�2�:�P��#�La WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�:�P��#�L��#N�#>b��t,>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�=�!��#�:� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CDV��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0v��HT�#��#$ұ��#4��D�D-��#�:B1�=B"Q��>�"��>�#6��L��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��# �H�#f�pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��@,`��D>�#>b��#>b��t,>b��=�HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��4� H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ;�!EB 4��#<� Q��# )a A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#ƎpH�P a��#ұ��#�:��a A,@�&!P�܈A� �J� �a� �A8 ֎PX�0�0H�b�CDƎP�2��@0��<� Qa A,@�&!P�܈A� �J� �a� �A8֎PX�0�PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ�:�P��#�L��# )�0HTa A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#&�@8��X<B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��֎PX�06� 8��#ұ�:�P��# �& )��@0��X<� Qa A,@�O6!P\Ď$��B$H0�� ֎PX�0�PHT�#6�@8��<B!��:�P��# )Ǝ�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d֎PX�0&� 8��#ұ;�!EB 4��#�,�m�CD��@0��<� Q��#�,a A,@�&!P�܈A� ��# �`�+.��!�&��A� �� ֎PX�0V�`��Է�#ұ�:�P��#�L��# )�0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d֎PX�0&� 8��#ұ�=�!�R߶�0T@d��@,`�H��#�<!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8֎PX�06� 8��#ұ�=�!�R߶�0T@d�H��# �& )�0HT`��,a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�>�!Q��v�@8��<BHLDH�2�:�Pa MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a FA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�PHT�#��`H��#��#(��D�D-Sm��t,>b�H��#$��D�D-��#�Hm�CDƎP�2a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#�Hm� 1!Q�d��#;B1�a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a JA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�PHT�#֏P8��Ə`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#N�#F��L��0T@a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��;�!EB 4��# ��Զ��H��#>b?��D>�#��#$��D�D-��#�=�!��a GA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ~�E�@֎PX�0F��L��#��:��b��D>�#��# ұ��6��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��?��D>�#��# ��Զ��H��#>b�<BHLDH�2�:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<wsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@֎PX�0V��L��#��:��b��4� H��#N�#>b��ܶ��HT�#>R?"��m� 1!Q�d�CDv�pH��#>a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CD��@0��X=B"Q��?B�D>r��>�"��>�#F��L��a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0��HT�#��#$ұ��#4��D�D-��#8��D�D-S�#>b�CDƎP�2��@0��X=B"Q��>�"��>�#a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a PA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#֏�8��>�#��0T@d�H��# �& )f��HT�#>�#Ə�H��#��#8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q�ě:��L��#��T5n��4U͛;BHLDH�2�<!8Me��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�Y<�!8MU��@NS�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#��T��L��#�<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2�&��4MF��4M��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��_�O�H#�d��4UM�<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!v��4M��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �q }2"�\�3�U<�^�3��>�`�3�I�b�48FM�d�4XVX�f�5x�Z�h�5��]�j�5�Fi�l�6�Vl�n�6��n�p�6�y�r�78G}�t�7XW��v�8x��x�8��z�8�G��|�9�W��~�9�G��:P��:8��:X��;x@��;�P��;�@Ƀ��<��̃��=�ۃ��>8pꃔ�?X`��?x��@��A��A�@(��B�@9��D�I��D8�X��EX�_��Fxo��G��~��H��I��J�𫄰�J��L80؄��Mh0߄��M��脼�N�0��N(��Nh0��N��N�0��O��OH0��O��O�@��O��PH` ��P��P��P ��PH��Q� ��Q��Q(0��QH(��Q�(@��R�(�(��R)P*��RH)�+��R�)P-��R�)�.��R*`8��SH*�9��S�*�;��S�*@=��S+�>��SH+@H��T�+�I��T�+@K��T,�L��TH,�N��T�,X��U�,�Y��U-0[��UH-�\��U�-`^��U�-�_��V.pi��VH.�j��V�.pl��V�.�m��V/�o��W(/py��Wh/�}��X�/p��X�/��X�/��X�/P��Y(80��YH8К��Y�8P��Z�8Ы��Z�8p��Z�8��Z9��[89��[x9p��[�9��[�9pȅ��\:Pʅ��\X:0̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<p��_�<��_�=P��`8>0��a�>)��bx?08��c�?�9��c�?`;��cH@=��cHH ?��d�HI��d�H�J��dI�L��dHI�N��d�I�X��e�I`Z��eJ@\��eHJ ^��exJ h��f�J@j��f�J@l��fK n��fHKx��g�K�y��g�K�{��gL�}��gHL��h�L`��h�L@��hM ��hHM��h�M��i�M��iN��iHN��i�N��j�Nઆ��jOଆ��jHO��j�O��k�O��kX`��kHX@��k�X Ȇ��l�Xʆ��lY�ˆ��lHY�͆��l�Y�φ��m�Y�ن��mZPۆ��mHZ�݆��m�Z�߆��n�Z醮�n�Z@ꆰ�n[`놲�n8[�솴�nX[�톶�n�[`�o�[��o\`��o(\��oH\��ph\��p�\��q�\��r�\�*��r�\�/��s]p<��t(]�O��uH]�i��vh]�|��x�]��y�]��{�]��|�]ه��}^��(^��ɀH^`��x^��^@)��^�,��^�8��_�<��8_�I��X_0[��x_Pl��_~��_��ʉ�_0��_��h��8hp��Xh��׏xh� ��Đ�h� ��hp��h��h�)��ip9��ړ8i�H��͔Xi�O��xi@[��Ε�i�^��i�n��i�}��i��j��8jP��Xj��xj��j`��j��j��j`��k��8k��ȟXk0��xk� ��ˠ�k��ȡ�k�(��k�/��k�<��l@I��פ8lpX��Xl�[��٥xlP_��l�j��Ŧ�l�o��ŧ�lP��l��m��8m ��Xmp��xmp��m��m��ݭ�m��m0��n`��ѯ8n��Xn@ ��xn� ��np��n@*��np9��ڳ�n�H��oL��۴8o`O��Xo`_��xo�n��o�}��op��¸�o ��oP��x��8x��Xx��xx��¼�x`��x��x��x��ǿyP��8y��Xy��xy0.��yP;��ypH��y X��y�[��z�_��8z�k��Xz�x��xz0��z��z��zp��z ��{��8{P��X{��x{ ��{@��{� ��{� ��{��|��8|0+��X|p:��x| I��|�O��|�[��|�_��|`o��}~��8}��X}��x}��}0��}0��}P��}P��~p��8~��X~`��x~��~ ��~ ��~��~P)��8��8�<��X�I��xpX��p\��h��l��y��8��X�P��x�`��`��؈��@�� 8��X��x��؉��0,��:��8��I��X��M��x��Y�� i��o��؊�~��`��@��8��X��]��.threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_61_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L&� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU�_~��DU!�!ȪR@X�)H�R��/@\�ȫR�@`��4@d�� ,��w�EiQE(c-b��(Bn�[�G��"E��]TYRK�"�(�"��E�A��/B�hF�E�c0"�)B ��a�HI��L�� c��RR#Ȩ�"T#MV�_�ʊ@�Pt S�a2#�(�"$�uF��2\р$Z��"�W�u�V�+�P+�k�A�� pE\~�ke��.E�2\q�%$[��+��W�% �V�+C�p+��A�� !��pEb��oe��L�2\�)$\��"�W$�W�+�"��+�� Cʕ� Β!��p�g�sS�T0�@ �1�@F�"Xg]W��+�u��vE�n-��E��-�w�Q��q�w�Q��"�`�]�^�Ӌ�3��"X�W�� a4_�K��+�u��}E��1ʯ�IF��_��_f�!~��_� g��!�~F�� q�Z@� �H�B\�w�Ÿ��;��r�W�a1��$,�}`QX�+��W@� �4,��q�w��1 �q�cb!��H�B\��/��Ÿ��_~��2�W|F9�",w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �Q��0;�s� �e�p�c0��`��q�b0�($r�8H1�`��`�@��\0S ��L.�)��'��v@~�Ƒ��@;��P�` �|�q�b0�(�r�8X1�`$v�`�@�ڿ\0W ��b��@;��r�8`1�`d�A.G,�~5��` ��`�@��c��v@��r�c0��Y@t ��x@ȑ��@;�qd9�1�`T��+G-�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0] �� Ʊ��@;��\�8x1�`TƂ+G/��Yp��` �Ph�`�@�zq��v@�.��q�c0�h�W0q �� 1��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]�K�H��|�@�Ʌ�0H�Ho�x�@��"�3��C��f�Z `�d�W�D�Y��!Rk3�|�0D�mfZ�H��X��i��U"�8�� C�;g�U `�d�̵*@<�z��!R�3�0Dy��H0�@�@� �)>"�=I�H�a��!�3�|0D�}��H��<�@�I�i>"�`#�C$l�� `��M4�@�ad��!+6��0D��F��H��D�@��V"�d�� C$�lZ `�T��C+�貑h��!�i6�0D��&��H��D�@�I?�V"�h�� C�m$[ `�t�Md+��ah��!�6�0D��H��<�@��]�V"�l#� C$�m"Z `�4�MD+�趙p��!��6��0D��f��H��X�@��W"�p�� C�:n0\ `��ͅ+�ȹ�p��!�>7�0Dr�� H<�8�@�I��W"�v�� C$�n.^ `��.�D��!R�7�0DZ��H��`�@�)�\"�{s�C$�o.p `��$.�侱��!�7��0D��H��L�@�I� \"��c�C�9t,p `��.��!R>:�0D�I��Hg�,�@�I.��\"��C$�ts `�4�Β.D�|�Y �H��,�@��>�]"��C$u"y `�T�NT/��y��!R�:Ͼ0D�U��H��,��i^�%]�"��@�R�Y��!�:Q�0D*['"�Hv�H'@� uIe"��C$�u � `��53�`�Yj��!�;��0D2c'��Hw�D=@� � l"�ÐC$svr `�$�c.D��a��!�J;L�0D�i��"��ÐC$�v"v `��N�.�`ۑ��!�x;R�0D�o��H$�@�@��Ɲ�_"��ØC�:w�t@`��0Dju�!�H��H!@�i��d"��3�C$�w&� `�4�N�2��މ`��!R�;��0D�}��H��07@��Ig"��#�C$|&� `��C.��a��!�(>��0DzŇ1"�H��0�@�� `�4�C.��!L>ѻ0D �G��Hs�H�@��4_"��C��|* `��c.H�L�a� �!��>@��G�\"��#�C$>}$� `��t2��T��!�>Q�0D*�'��H6�\3@��m�f"��C��}$� `�t��3��!�>�0DB�!�H4�0�@� ��\�"��äC$<~��!)?�0D��'b�H��D�@�I�I^"��#�C$�~ | `�d��/��!�j?�� 0D��I$�H��0C��~r `��ߏ2�Z��H��!ҏ?��0D��gR�H��D-@��֟f"��s�C$�� `��s3�l��t��!�?R�0D��g�H��0�@�)��\"��ØC��s `��.D�Rb �H51�@�)��]"�!�C�Q�$y `�T��T/��!�=Bо0D:I��H< 1�� -!&]�"�%�@�@b��!kBR�0D�MH"�H� M'@�iD�Ie"�(�C�3�"� `��53�^bj��!ҮB��0D:WH��H� I=@�ie� l"�,ĐC$��r `�d�c.D�Db��!��BL�0D�]��"�/ĐC$��"v `�4�P�.�h��!�CR�0D�c��H�A�@� ��_"�3ĘC�|��t@`�T�0D�i�!�H= I!@��!�d"�64�C$چ&� `�t�P�2��`��!R�C��0Ds��Hl17@� �!Ig"�:$�C$^�&� `��#6��b��!��C�00D�y�!&�H?1f��i�!m"�=Ĉ C�ч$� `�T��6��!��C�0Dj�(��H>Et@�� #�n"�bĐ C�X��`�ԋ�6��1b��!R=F�0DB�h��H;Q|@��-#�o"�f$�C�ڌ$� `�t�F:��3bL��!ҊF�0D��hR�HjM�@��O#�u"�jĈ C�^�� `��C6L��5b�� !R�F��00DjوA�H91b@��l#Im"�m$� C�ԍ&� `�Ļ�7��7��!G�0D��Ȳ�Hg1d�i�#�l�"�qĠ C�U�� `�t�ф7��9��!R?G�0D��HeI�@��#It"�v��C�� `��:��;�T��!�G�0D��r�Hs1b@��#Fl"�zĐ C�[��``�4�c6L��=b��!ҾG��0D:�HR�H9Il@�i��m"�}4� C�؏"� `�$�QD7��?��!R J�@0D�A�1(�Ha(1h@�� %Fl"�4� C$V�&� `��7��Q��!�IJ�0D�II�Hk)a�@�i4%�t"��ĠC�ݔ&� `��:�BT�\��!R�J��0D S��Hm*1d��)O%�l�"��Ę C�X�� `�ԫ#6��U��!ҾJ�0DjYir�H`+Mp@�in�Hn"�� C�ڕ,� `��C6P��Wb� �!�K�0Dza��H;,Mx@�ɍ�Io"��D� C�Z�(� `��Β:��Y�D��!RKK�0D*k�1�Hq-1�@��Iu"��4�C��(� `�tߒ5�Z\��!�K�|0D�s�$6��$2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B� b D�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�align�+0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1��..6�//6�0/06�`/16��/26��/36��/46� 296�P2:6��2;6��2<6��2=6�3>6�@3?6��3b6��3c6�6d6��.l6�/m6�0/n6�`/o6��/p6��/q6��/r6� 2w6�P2x6��2y6��2z6��2{6�3|6�@3}6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.&7�/'7�0/(7�`/)7��/*7��/+7��/,7� 217�P227��237��247��257�367�@377��3:7��3;7�6<7��.d7�/e7�0/f7�`/g7��/h7��/i7��/j7� 2o7�P2p7��2q7��2r7��2s7�3t7�@3u7��3x7��3y7�6z7��6�7��a0��6�7��a0��6�7�7�7��a0�07�7�P7�7��b0�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6 � �M#s#��B݌`0�p7$� �M #{��B߬`0��7,� �N#��B�`0��:2� ��#�C��`0��6=� ��#q��܀a0�P7a� ܍#ys��ޠa0��7i� ��#��a0�P:m� �#��a0� 3D �0�0�`�3Ip�0(�1ð�C03N��0<3��D3S�0PA4�P��D3W��0`A4Ð��D3[�0pA4��D3_3a��`0�0�0�A3e��`0�p�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�A��0�A�0�A�0�A �0�Ap�0�Ap�0�A��0�A��0�A�0�A�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�B�0�B�0�B��0�B�0�B�0�B �0�Bp�0�Bp�0�B��0�B��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�C��0�C��0�C�0�C�0�C��0�C�0�C�0�C �0�Cp�0�Cp�0�D��0�D��0�D�0�D�0�D �0�Dp�0�Dp�0�D��0�D��0�D�0�D�0�D��0�D�0�D�0�D�30��L11�0M�0�D@36��N71Ð�K�0�D3;��O41��M�0�Dp3?��X41��M�0�Ep3c��Y11�PAL�0�E@3g��Z71Ð�M�0�E�3k��[11��M�0�E@3o��\71�M�0�Ep3s��]/1�PAL�0�E3w��^41Ð�M�0�Ep3{��_11��AL�0�E@3��h71��M�0�F@3��i71�P�K�0�F3��j41ÐM�0�Fp3��k/1��AL�0�F3��l41��M�0�Fp3��m71�P�M��h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n��H`��q��kK�s�+��K{s%��\��FQ ۸ �� (�8�H�X�h�x�� )9IYiy��0 12)394I5Y6i7y8�9�:�;�<�=�>�?�P QR)S9TIUYViWyX�Y�Z�[�\�]�^�_�p qr)s9tIuYviwyx�y�z�{�|�}�~�� )�9�I�Y�i�y�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� d��\Ǝ��4Ma A,$�� d��\Ǝ��4Ma A,T�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,�� 6C�A8 ��\��`��4M֎��4Ma A,D�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,t�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� d��\Ǝ��4Ma A,d�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,D�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,t�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,4�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,4�� 6C�A8 ��\��`֎��4M��4Ma A,d�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� d��\Ǝ��4Ma A,�� d��\Ǝ��4Ma A,$�� 6C�A8 ��\��`֎��4M��4Ma A,T�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma A,T�� d��\Ǝ��4Ma A,�� 6C�A8 ��\��`֎��4M��4Ma A,�� 6D0�B�@ ��\��`Ǝd��4M��4M��4Ma A,�� A�6E0��lCP��4M��4M֎X��\��`Ǝd&��4M��4Ma Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@֎0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1֏`NSٽ &$@#��&ұ<� �&$@#��#��Twm� ��4Ս�?� 0��#�,6o��V�`NS�&��T8m� ��<4n��4�m��4Un��4o��4��#��T5n��4U��:B�:B=!�n� ��<wn��4�M�>�!8Meۆ�@��>�!8Me�a �� G�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|ğ\p��A|ԟ��A|ܟ\p�@�@T� WC��5֏`NSٽ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D&�@̀[<B!8Me�<B!8M�>B!Q}[=B!8M��;!8Me��4U][>�!8MU��`��ܷ&��Txo� ��4U��:B�:B;B =!�n� �,na�!˂�f�`NS��#@3Դ�#��T�m��4��?!�a �� { �� G{B�@��Ǝ֎`��#�,>Ra �� z A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|ğ\p�@@��A|ԟ�@��A|ܟ\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�%&Nd� ��4��`�!�r�&�@̀[<B!8Me[=B!8M��<B!8M��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}?!�a �� G� �� ć�B�@��Ǝ�P��֎@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5֏`NSٽ�#�,��`NSٶ!&ұ<� ��#��T7n� ��4խ[`�!8Mu��0�`� ��4�[b��D�PNS�V�`NS��#��T��#@3��PHGT��@NSY=B!8M��=�!8MUז�`NSո&��T�m� ��4U��:B�:B;B =!�n��,�}�`�!˂�F�`��#@3Դ�#��T6a �� ć �� B�@��Ǝ֎`��#�,>Ra �� G A�@��Ǝ@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5��`NSټ!&ұ<� �&��T7m� ��4�m�`�!8Mu��`��ػ�#@3Xa�!8M��f�`NS�&��T�n� �,˭�<4n��4�m��4Un��4o� �>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}?!�a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|ȟ\p��A|؟��A|��\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ �@N3�;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��AƘ��0H�b��tD��0�`��4�m��4Un��4o��,�}<!8Me��4U]�>�!8MU㦎��6�`��න#�&�,8o��5m��4�m[?!�b��4��?�!8Me��`NSݸ&��T�n� ��<&�`NSݼ-&��&��T8n� ��a�!8M��6�@̀[>B!Q}�a�!8M�צ�`NS��#��֘`NS�9&��T� a � ��A|�_ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|�\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m� �,��<B!8M�[=B!8M�>B!Q}�;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�^ d��Ǝ֎0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<��`NSټ1&ұ<� �&��T�m� ��4Ս�`�!8Mu�V�`NSݼ&��Two��v�`NS�%&��T�n� ��<4n��4�m��4Un��4o��tD�m��4��#��Tum��4U��b�!8M�ۦ�`NS崩#��#��#��#�톘`��#�0aBý�=B4CM�>�!8Me۶�`NSٸ�#��Tvn��4��`!�a A ��A|�W d��Ǝ֎0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<&�`NSټ1&ұ<� � &��T7m� ��4�m�a�!8Mu�v�`NSݹ�#@3Xb�!8M�Ӧ�`NS�-&��T8n� ��<4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U][>�!8MU㦏`NSպ�#��T5o��4Uݛ:B�:B;B =!�n�0F�P�p�f�P�P��`NSٴ�#��T�m� �,�&��T�a � ��A|�_ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me[>B!Q}=B!8M�<!8Me��4U]�>�!8MU�f�PNS�`NSἩ#��#��#��#�햘`��9&��Tym� ��4U�[?��I&��Tyo� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a � ��A|�^ 7x`��@@� d ��0�c�`�A`��`�0�@N3�;!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H�`NSټA&$@#��M&ұX<� �5&$@#�� &��T�m� ��4Սa�!8Mu�Ƙ0�c� ��4��a�!8Mu��P̓��0�`� ��4^�b�!8M��6�@̀�<B!8Me=B!8M��=B!8M�[>B!Q}<!8Me��4U]�>�!8MU��`NS�-&��T8o�`�A`��`� ��4U�[=!�n� �,�[d�!8M��֏0�`� 0�n��5m� ��>�!8Me�Ə`NSٸ�#��Tvn��4��[`!�a �� p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� { �p1��Q�D8�0�c�`�A`��#��T��#��T֎`��a �� G� �p1��Q�D8 �0�c�`�A`��#��T��#��T֎@��a �� ć �p1��Q�D8�0�c�`�A`��4�n��#��T֎`��a bA,@��Uz�P��w@!�^p��x�pr2b` ��~Bp�p r2b` �Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` ��~Bp�p q1b` �Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<�`NSٽ&��-&ұX<� ��#��Twm� ��4�}�`�!8Mu��0�c� ��4�͛`�!8Mu�֏`��ؼ�#@3Xa�!8M��v�`NS�!&��T�m� ��<4n��4�m��4Un��4o��4��#OD��=�!8MU㆏`NSպ%&��T8o�`�A`��ÿ8V�@HG��b�!8M�Ӧ�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽ-&ұX<� � &��T7m� ��4�]a�!8Mu�V�`NSݺ&��a�!8Mu�֏`��ظ!&��Two��`NS�)&��T8n� ��<4n��4�m��4Un��4o��4��#��T5m��4Um;��D��#��T�n��4U͛:B�:BX;��D��/��#��f�P�s�Ə`NSٴ�#�,�`NSٹ�#��T�a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��@��a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� H� A�@��Ǝ@��a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a �� A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� �� #�,a �� H� A�@��Ǝ@��a �� ȿ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a �� Ȼ A�@��Ǝ@��a ��A� �� #�,a ��A� �� #�,a ��A� ��@��a ��A� 䇹��@��a �� H� A�@��Ǝ@��a �� ȸ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� Ȭ A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ȩ A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� Ȩ A�@��Ǝ@��a �� B�A8��Ǝ0@�c��,a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� Ȉ��@��a ��A� ��@��a �� ț A�@��Ǝ@��a �� Ј� B�A8��Ǝ֎@��a �� Ț A�@��Ǝ@��a �� H� A�@��Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� ؈� B��@��@H�;!<Sa �� B��@��@H�;!<Sa ��A� ��@��a ��A� ��@��a �� H� A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a �� ȋ A�@��4Ǝ@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� A�@��Ǝ@NSY;!�a � �� B�@��Ǝ@NSY;!�a �� G A�@��Ǝ@��a �� G A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� I~ A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� } A�@��Ǝ@��a �� H� A�@��Ǝ@��a �� { A�@��Ǝ@��a � �� {B�@��Ǝ@NSY;!�a �� { A�@��Ǝ@NSY;!�a �� z A�@��Ǝ@��a �� z A�@��Ǝ@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��T��@��<BS,�Oa �� y A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Iy A�@��Ǝ@��a �� Ț A�@��Ǝ@��a �� o A�@��Ǝ@��a � �� HoB�@��Ǝ@NSY;!�a �� Ho A�@��Ǝ@NSY;!�a �� n A�@��0T�b�0!IƎ@��a �� m A�@��0T�b�0!IƎ@��a ��A#�` &�5��0T�b�0!IƎ04�a��,a ��A#�` &�5��0T�b�0!IƎ04�a��,a �� l A�@��0T�b�0!IƎ@��a �� l A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� k A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� j A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� i A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� h A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� _ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ^ A�@��0T�b�0!IƎ@��a �� ] A�@��0T�b�0!IƎ@��a �� ,u5A�@Ǝ0T�b�0!I��p�a��,a �� ,s5A�@Ǝ0T�b�0!I��p�a��,a �� \ A�@��0T�b�0!IƎ@��a �� [ A�@��0T�b�0!IƎ@��a �� I[ B�A8Ǝ0T�b�0!I��p�a��,a �ȈA� ��Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a �ȈA� � Z� ��@�@ ֎0�c��4��#��T�PNS��PNS�Ǝ@��a ;J,@ �� a��pC�`0�0�� [%�1K�7�2F0bp p�V��#�`'m�,�,�1PAD!��@y(�p�$��A86�`��Խ�#�<B!0OM>�!8̀Ӗ�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#f��p�Q��,Ca#�`�#g�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3v�`N3ؼ�#��@̳>�!8�`�F�`��ص�#�:B�:BX=�!�b�6�P�S��@��a 9J,@ �� Q�$� E��,�@#�`�#a��p�Q��,Ca#�`�#b�$�ǈ�� X-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��>�!8�`��@��a 9J,@ �� Q�$� E��,�@#�`�#<��p�Q��,Ca#�`�#=�$�ǈ�� HO-�,�1PAD!��@y(�p�$��A8&�`��Է�#�3f�@N3X<!0�n��,�]�;� 0��#��#��#�,�n��<un��4��<!�b��4��a 5J,@� ��,C #�`�4E�$� ��,�@#�`�5e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�/E�$� ��,�@#�`�0e� � F��~�2Kp��~�T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'*E�$� ��,�@#�`�'+e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 5J,@� ��,C #�`�'%E�$� ��,�@#�`�'&e� � F��2Kp��T2KpTD!��@��1K�TP �֎0�c�`��4��#�3V�`N3Զ�#��m��,Kn��<��#�,��#��@��<�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP |2R�,�1bP |2R�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8�0�c�`��4��#�3f�`N3Զ�#��m��,Kn��<��#�,��#��@��=�!�R�a � �X�@��@0�@;!8Ma � �T�@��@0�@;!8Ma � �`�@��@0�@;!8Ma �� I< A�@��0T�b�0!IƎ@��a �� ; A�@��0T�b�0!IƎ@��a �� I; A�@��0T�b�0!IƎ@��a �� : A�@��0T�b�0!IƎ@��a �� I: A�@��0T�b�0!IƎ@��a �� 9 A�@��0T�b�0!IƎ@��a �� I9 A�@��0T�b�0!IƎ@��a �� 8 A�@��0T�b�0!IƎ@��a �� I8 A�@��0T�b�0!IƎ@��a �� / A�@��0T�b�0!IƎ@��a �� I/ A�@��0T�b�0!IƎ@��a �� . A�@��0T�b�0!IƎ@��a �� I. A�@��0T�b�0!IƎ@��a �� - A�@��0T�b�0!IƎ@��a �� I- A�@��0T�b�0!IƎ@��a �� , A�@��0T�b�0!IƎ@��a �� I, A�@��0T�b�0!IƎ@��a �� + A�@��0T�b�0!IƎ@��a �� I+ A�@��0T�b�0!IƎ@��a �� * A�@��0T�b�0!IƎ@��a �� J* A�@��0T�b�0!IƎ@��a �� ) A�@��0T�b�0!IƎ@��a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a ��A#�` &�4��0T�b�0!IƎ04�a��,a �� I( A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� ,*4A�@Ǝ0T�b�0!I��p�a��,a �� ,(4A�@Ǝ0T�b�0!I��p�a��,a �� ,&4A�@Ǝ0T�b�0!I��p�a��,a �� ,$4A�@Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� B�A8Ǝ0T�b�0!I��p�a��,a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �� I�A�@��0T�b�0!IƎ@��a �� A�@��0T�b�0!IƎ@��a �ȈA� �I��@NS�:�D�`��,a �ȈA� ��@NS�:�D�`��,a �ȈA� �I�� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a �ȈA� �� @�@ ֎0�c��4��#��T��#��T��#ODƎ@��a ��A� � ��0<Sa ��A� ��0<Sa ��A� � ��0<Sa ��A� ��0<Sa ��d�@��<a ��d�@��<a ��d�@��<a ��d�@��<a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a �� ̇�B�A8Ǝ`8E!��# )��@�Da ��A� �� a �� J��0�d�2a ��A� ��@0��a �pCPa0bp ��6�� Ap�A8��@<1�:��a A,@ � �� B0��@<1�:��a �pCPa0bp ��6C� Ap�A8��@<1�:��a A,@� �� B0��@<1�:��a A, 4�\W ��8�C:�C9�C8��:��8́��} #D�` a &A,@��3��<��9��9��A��6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa 'A,@��3��<��9��9��A��6!P�\Ď��| dƎPX�0�PHT�#��pH�P a�CD&�@8��X<B!��a #A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa $A,@=s�A��:�;�A=��A��&!P�܈�Q� ��pA�@ ƎPX�0��PHT�#��0T@d��@,`��t,>a %A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��#�L��#a 'A,@��3�<��9��;܃;��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��@,`��#ұ�a &A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa 'A,@��3�<��9��9�C>��8́�� } l l l 6!P�\Ď��| dƎPX�0�PHT�#��0T@d��4� H��#N�#�PH��#a &A,@�M5��<��9��9��A��6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�!EB 4��#<� Qa #A,@�Ts�A��:�;�A=��A��&!P�܈�Q� ��pA�@ƎPX�0�0H�b�CD��P�2��0HTa %A,@�M5�<��9��;܃;��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��#�L��#a &A,@�M5�<��9��9�C>��8́�� }`l`l l 6!P�\Ď�� d ƎPX�0&� 8��#ұ�:�P��# �& )�0HTa UA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��m� 1!Q�T[>�!Q��V�PHT�#v�@8��>�!��f�PH��#6��L��0T@a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a OA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�PHT�#v�`HT�#��#(��D�D-Sm��t,>b[=B!��<BHLDH�2�:�P��#�La QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#��P8��ܦ�`H��#��#(��D�D-Sm��#�Hm��t,>b� 1!Q�d�CD��P�2a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��#ұ��<�HLDH�2��# �H�#a UA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�PHT�#F��L��#ұ��m��#N�#��#�Hm��t,>b� 1!Q�d�CD��,@C��a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#��HT�#��=�!��#,��D�D-��a WA,(@��3��<��9��9��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m�� %��HT�#>R=�HLDH�2զ��H��#>b[=�!Q��#ұ��=��D>�#6��L��0T@a QA,'@=s�A��:�;�A=��A��@?��T;��=��8��>�A��A?��@��T;��=��8��>��A?��} l@#E�`�;�1�� B`��| ��@~p#G�`0��#F�`�'K@dƎPX�0v��HT�#>R>"��m� 1!Q�T=�!Q��#ұ��<BHLDH�2�:�P��#�La WA,(@��3�<��9��;܃;��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�:�P��#�L��#N�#>b��t,>�#a WA,(@��3�<��9��9�C>��8́�� } l l l ��\]@��;؃��}@l l ��\^@��;؃��A��} l��#E�`�;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�'MPdƎPX�0��p8��m��D>�#��# ұ��F��L��#��<BHLDH�2�=�!��#�:� a�b��a XA,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d��D>R��#N�#>b��t,>�#6��L>a [A,(@��3��<��9��9��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CDV��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a QA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0F�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`��D>R��#ұ��<�HLDH�2�a TA,'@=s�A��:�;�A=��A��@��T;��=��8��>��A?��} l��\]@��;؃��}�l l #E�`p;�1�� B`��| ��@~p#G�`0��#F�`�;J@dƎPX�0v��HT�#��#$ұ��#4��D�D-��#�:B1�=B"Q��>�"��>�#6��L��a WA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��P�2v�`8��# �H�#f�pH��#>b� 1!Q��#a ZA,(@��3�<��9��;܃;��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��@,`��D>�#>b��#>b��t,>b��=�HLDH�2�>�#a XA,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CD��,@C��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a [A,(@��3�<��9��9�C>��8́�� } l l l ��\^@��;؃��A��} l��@��T;��=��8��>�A��A?��#E�`p;2�� B`H��U��Q� �GB4�1MT WA�#G�`0M#F�`�;LPdƎPX�0��HT�#��# N�#>�#f��H��#>�#6��L>�#��0T@d��4� H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ;�!EB 4��#<� Q��# )a A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#ƎpH�P a��#ұ��#�:��a A,@�&!P�܈A� �J� �a� �A8 ֎PX�0�0H�b�CDƎP�2��@0��<� Qa A,@�&!P�܈A� �J� �a� �A8֎PX�0�PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��L��&@�@֎PX�0&� 8��#ұ�:�P��#�L��# )�0HTa A,@�&!P�\Î$��L��&@�@֎PX�0�PHT�#&�@8��X<B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��֎PX�06� 8��#ұ�:�P��# �& )��@0��X<� Qa A,@�O6!P\Ď$��B$H0�� ֎PX�0�PHT�#6�@8��<B!��:�P��# )Ǝ�,@C��a #A,@�&!P�\Î�A~�2b� ��2J�0�#�`�+'d֎PX�0&� 8��#ұ;�!EB 4��#�,�m�CD��@0��<� Q��#�,a A,@�&!P�܈A� ��# �`�+.��!�&��A� �� ֎PX�0V�`��Է�#ұ�:�P��#�L��# )�0HT`��,a "A,@�&!P�\Î�A~�2b� ��2 �0�#�`�+'d֎PX�0&� 8��#ұ�=�!�R߶�0T@d��@,`�H��#�<!�a #A,@�O6!P\Ď�A~�L2b� �3K$�hB��L�A8֎PX�06� 8��#ұ�=�!�R߶�0T@d�H��# �& )�0HT`��,a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�>�!Q��v�@8��<BHLDH�2�:�Pa MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a FA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�PHT�#��`H��#��#(��D�D-Sm��t,>b�H��#$��D�D-��#�Hm�CDƎP�2a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�PHT�#ƏP8��ܶ�`H��#��#(��D�D-Sm��t,>b�H��#�Hm� 1!Q�d��#;B1�a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a JA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�PHT�#֏P8��Ə`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#N�#F��L��0T@a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��;�!EB 4��# ��Զ��H��#>b?��D>�#��#$��D�D-��#�=�!��a GA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ~�E�@֎PX�0F��L��#��:��b��D>�#��# ұ��6��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��q7�P��`t�@p�d��`@r�G��<�9#E�`�;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�'O`d֎PX�0F��L��#��:��b��?��D>�#��# ��Զ��H��#>b�<BHLDH�2�:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<wsy��`n v@`�dl@q�ld {p�<�A��#E�`�;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ~�F�@֎PX�0V��L��#��:��b��4� H��#N�#>b��ܶ��HT�#>R?"��m� 1!Q�d�CDv�pH��#>a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0ƎpH�P a��D>�#Əp8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#6��L>b��D>R�a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3�)(Q0�#F�`�;N`d֎PX�0��HT�#��#R$�@C�=B"��<BHLDH�2��=�D>�#>b�CD��@0��X=B"Q��?B�D>r��>�"��>�#F��L��a HA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0V�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��$��L�K0��0!P(܈A� ��D�� 0b`T ��E�@֎PX�0��HT�#��#$ұ��#4��D�D-��#8��D�D-S�#>b�CDƎP�2��@0��X=B"Q��>�"��>�#a MA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0V�pHT�#>b��#��H��#��=HLDH�2�>b�CDƎP�2��@0��=��D>�#f�pH��#>b� 1!Q��#��HT�#��a PA,@�a7�P��`t�@p�d��`@r�G��<�;�vsu��`n v@`@p�`�dl@r�{`�;��<��<#E�`p;�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� ��3S)(Q0�#F�`�;N`d֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0f�pHT�#>b��#Ə�H��#��X=HLDH�2�>b�CD��@0��;� a�b��=�!��#,��D�D-��>"Q��>a PA,@�Ovsy��`n v@`�dl@q�ld {p�<�A��i7WP��`tdl�@q�d ��F��#E�`p;2�� B`H��U��A� � �< �� `��U��A� �J�P�"� 0b`T ��F�@֎PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#֏�8��>�#��0T@d�H��# �& )f��HT�#>�#Ə�H��#��#8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q�ě:��L��#��T5n��4U͛;BHLDH�2�<!8Me��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,T�y7�a�� n0�� n0x�@p�d��`�d ��G��<�;#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#E�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#E�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#E�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#E�`@.��A8��0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a KA,D��7�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#�Y<�HLDH�2�&�@NS<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a .A,��7b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,��7h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��7b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,D��7h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T��7b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#E�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#E�`p3��A8��0<Sa .A,4��7b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a AA,�� 8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�Y<�!8MU��@NS�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#��T��L��#�<�HLDH�2�6�`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma 2A,��7�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2�&��4MF��4M��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,T�y7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�;�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��_�O�H#�d��4UM�<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��4�m[>B��>�HD4�=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�``>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�>�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�>�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#E�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#E�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#E�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#E�`@.��A8��0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#E�`�>�0�� Bpp0�p�lC �掐�L��L��#�<!8Me��4UM�<B��>�HD4�<B��>�HD4a OA,D��7W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#E�`?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!v��4M��L��#(��D�D-�n� 1!Q��<�!8MU㦎0<Sd��4U͛;BHLDH�2�=B��>�HD4�<!8Me��_�o�H#�d��_��H#�d��4�ma 2A,��7�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#E�`0?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,��7�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#E�`P?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��7�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#E�`p?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,D��7�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#E�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T��7�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#E�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#E�`p3��A8��0<Sa 2A,4��7�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#E�`�?�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a EA,�� 8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#E�`B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y<�HLDH�2��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#E�`@B�0�� B0�P�lC �掐�L��L��#�<B��>�HD4Y<B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #G�`�B�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #I�`�B�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#G�`�BC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�B�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#G�`CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�` C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#G�`@CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#F�``B`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#F�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#F�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#F�` 6� `8��0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#G�``C�!�0U�B~!~��d��4n��4��#�Y;�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#I�`�C�BB��@��0U�B~!��d�S,��>qM L&�PNS�6�PNS��@NS=�!8MUӦ�0<Sd�E,a�S,��>qM L�0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#I�`�C�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#G�`�CC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#I�`F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#G�` FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#F�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#F�`�:� `8��0<Sd�E,a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#G�`@FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a ;A,t�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#I�``F�AB��@�0U�B~!��d��3E��PT�B֎0U�B~!>��d�S,��>qM L��0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#G�`�FC!��0<Sd�S,��>qM L�0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#G�`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#G�`01�� #�c��t �q }2"�\�3�U<�^�3��>�`�3�I�b�48FM�d�4XVX�f�5x�Z�h�5��]�j�5�Fi�l�6�Vl�n�6��n�p�6�y�r�78G}�t�7XW��v�8x��x�8��z�8�G��|�9�W��~�9�G��:P��:8��:X��;x@��;�P��;�@Ƀ��<��̃��=�ۃ��>8pꃔ�?X`��?x��@��A��A�@(��B�@9��D�I��D8�X��EX�_��Fxo��G��~��H��I��J�𫄰�J��L80؄��Mh0߄��M��脼�N�0��N(��Nh0��N��N�0��O��OH0��O��O�@��O��PH` ��P��P��P ��PH��Q� ��Q��Q(0��QH(��Q�(@��R�(�(��R)P*��RH)�+��R�)P-��R�)�.��R*`8��SH*�9��S�*�;��S�*@=��S+�>��SH+@H��T�+�I��T�+@K��T,�L��TH,�N��T�,X��U�,�Y��U-0[��UH-�\��U�-`^��U�-�_��V.pi��VH.�j��V�.pl��V�.�m��V/�o��W(/py��Wh/�}��X�/p��X�/��X�/��X�/P��Y(80��YH8К��Y�8P��Z�8Ы��Z�8p��Z�8��Z9��[89��[x9p��[�9��[�9pȅ��\:Pʅ��\X:0̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<p��_�<��_�=P��`8>0��a�>)��bx?08��c�?�9��c�?`;��cH@=��cHH ?��d�HI��d�H�J��dI�L��dHI�N��d�I�X��e�I`Z��eJ@\��eHJ ^��exJ h��f�J@j��f�J@l��fK n��fHKx��g�K�y��g�K�{��gL�}��gHL��h�L`��h�L@��hM ��hHM��h�M��i�M��iN��iHN��i�N��j�Nઆ��jOଆ��jHO��j�O��k�O��kX`��kHX@��k�X Ȇ��l�Xʆ��lY�ˆ��lHY�͆��l�Y�φ��m�Y�ن��mZPۆ��mHZ�݆��m�Z�߆��n�Z醮�n�Z@ꆰ�n[`놲�n8[�솴�nX[�톶�n�[`�o�[��o\`��o(\��oH\��ph\��p�\��q�\��r�\�*��r�\�/��s]p<��t(]�O��uH]�i��vh]�|��x�]��y�]��{�]��|�]ه��}^��(^��ɀH^`��x^��^@)��^�,��^�8��_�<��8_�I��X_0[��x_Pl��_~��_��ʉ�_0��_��h��8hp��Xh��׏xh� ��Đ�h� ��hp��h��h�)��ip9��ړ8i�H��͔Xi�O��xi@[��Ε�i�^��i�n��i�}��i��j��8jP��Xj��xj��j`��j��j��j`��k��8k��ȟXk0��xk� ��ˠ�k��ȡ�k�(��k�/��k�<��l@I��פ8lpX��Xl�[��٥xlP_��l�j��Ŧ�l�o��ŧ�lP��l��m��8m ��Xmp��xmp��m��m��ݭ�m��m0��n`��ѯ8n��Xn@ ��xn� ��np��n@*��np9��ڳ�n�H��oL��۴8o`O��Xo`_��xo�n��o�}��op��¸�o ��oP��x��8x��Xx��xx��¼�x`��x��x��x��ǿyP��8y��Xy��xy0.��yP;��ypH��y X��y�[��z�_��8z�k��Xz�x��xz0��z��z��zp��z ��{��8{P��X{��x{ ��{@��{� ��{� ��{��|��8|0+��X|p:��x| I��|�O��|�[��|�_��|`o��}~��8}��X}��x}��}0��}0��}P��}P��~p��8~��X~`��x~��~ ��~ ��~��~P)��8��8�<��X�I��xpX��p\��h��l��y��8��X�P��x�`��`��؈��@�� 8��X��x��؉��0,��:��8��I��X��M��x��Y�� i��o��؊�~��`��@��8��X��]��.threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_62_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z6__shfliii_Z6__shfljii_Z9__shfl_upiji_Z9__shfl_upjji_Z11__shfl_downiji_Z11__shfl_downjji_Z10__shfl_xoriii_Z10__shfl_xorjii_Z6__shflfii_Z9__shfl_upfji_Z11__shfl_downfji_Z10__shfl_xorfii_Z6__shflxii_Z6__shflyii_Z9__shfl_upxji_Z9__shfl_upyji_Z11__shfl_downxji_Z11__shfl_downyji_Z10__shfl_xorxii_Z10__shfl_xoryii_Z6__shfldii_Z9__shfl_updji_Z11__shfl_downdji_Z10__shfl_xordii_Z6__shfllii_Z6__shflmii_Z9__shfl_uplji_Z9__shfl_upmji_Z11__shfl_downlji_Z11__shfl_downmji_Z10__shfl_xorlii_Z10__shfl_xormii_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L Q&ӄAB� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU H�b<QU�_�Ȫ"�aU 0H�R�9@\��ȫ"PP�Ew��h,*�e�EfQE�-J��(�k[�Hq��"K�`I\$�R�P袺H�"Hs�Eh �h/��t�E~Q E0 �)B#�a�I21�#R�@Jb�US�j�Ɗ�KB�X��"�a�0�dD%S�d,��"��iV�+0�8+�P@�� p��he��2��2\�VdZ�X�B�W��V�+ⲄX+�t�B��ኻ�!��p�^h�le��/:��2\X d[��W� �V�+#�x+��UC��ኄ~!��p�P�pe�b�a��2\aY2d\��B�`)W��+�C�P��+�S@�P��+�S@�~Q��+��iW�;ĺ��n ` `� �a�"��P��`�yE �J��E��E��;��Z�!��;�0�[D_ȋ��+B{t~F0��"�w�W��^_�_��,�|�рE �0*��%FA?�(A#��K��4�/��K��49�!��H�2\QDa!�(��b\�aX�+ֺ�ҰW�u�w�A⊸�<,�t��. q�^@$㊾��B\�w�Ÿb�;��R�W�b1�H�\,��`�X�+&��WX��/��B4��!�l,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �!��0;��q� �e�p�c0��`��q�a0�($r�8B1�`��`�@��\0�Q ��L.)��'��v@~��a��@;��8�` �|�q�b0�(�r�8R1�`$v�`�@�ڿ\0�U ��b��@;��r�8Z1�`d�A.�+�~5��x�` ��`�@��v@��r�c0��Y@�r ��x@�a��@;�qd9�1�`T��+�,�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0�[ �� Ɓ��@;��\�8r1�`TƂ+�.��Yp��` �Ph�`�@�zq��v@�.��q�b0�h�W0�_ �� @;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]��H��t�@��/U"�1� C$Qf� U"�3� C��f�Z `��LW��j��!�n3\-0D�o�R��!�3dP�@�Y��!Ҫ3�|�0D�ufZ�H��X��)��U"�<�� C��g�U `�D�̵*@<�v��!��3�0D��H.�@�@��)>"�aI�D��d��!R-6�0Db�f��H��X�@��&KV"�e�� C��l0Y `��%+�䳹d��!R�6��0D�� Hj�8�@� O�V"�j�� C$sm.[ `�ԯͤ+�`��t��!��6��0D��H��`�@��|LW"�os� C$n.] `��+�@��x��!�-7�0Dr��H%�L�@� ��IW"�uc� C��n,] `�t� �+�޻�t��!҉7��0D��Hu�,�@� ԛ�W"�z��C$[op `��2.D�ؽY �H��,�@�I��\"�|�C$�o"v `�4�M�.�H�y��!��7O�0D��H��,��)�e\�"��@�n�Y|��!�:Ѿ0D�C'��H��H!@��d"��C$|t � `��2�|�Y^��!RZ:K�0D�K'��H��D7@��6Ig"�� C$�t_ `��.D�X�a��!�:̸0D"S��"�� C$>u"s `�ԪN�.�|Ց��!�:һ0DY��H2�@�@�Il� _"ݭÀC��u�q@`��0D�]��H��H�@�)�_"��3�C$v&� `��Nt2�DىT��!�,;W�0DRg�y�H��01@�I��f"Ѵ#�C$�v&� `��+�J�a|��!Rl;�0D�m�"�H��0�@�� `��+�Z܉��!��;Q�0DzsGb�H��H�@�i�H^"��C�yw*| `��.H�h�a� �!�;@��W"�#�C$�w$ `��2��ߙH��!�>��0D��'R�Hd�\-@�i�e"��C�Y|$� `�T��t3��t��!�J>��0D�ɇ��Hb�0�@��-\�"��ÌC$�|��!�l>��0D��'2�H��D�@� D�]"��#�C$2} y `�D�T/�H��!R�>� 0D*ׇ$�H��0C�~}_ `��/�v��!R�>S�0D�g"�H��D'@��|�He"��s�C$~�� `��3��h��!�?��0DB�g��H��0�@��W"��ÀC�~p `��3.D�n�a �Hc�0�@�魟�\"��C��~$v `�4ޏ�.��!R�?P�0D��Hj�0��f\�"��@�\�a|��!��?Ҿ0Db�G��H��L!@�)柉d"��C��"� `�d��2�z�a^��!R�?L�0D��G��H'I7@�)�Ig"�!�� C$<�_ `�D�.D�`b��!/B̸0DBG��"�#�� C$��"s `��P�.��!]Bһ0D2M��H� A�@��<� _"�'ĀC��q@`�4�0D�Q��Hk I�@��O!�_"�*4�C$q�&� `�T�Pt2�L�T��!��BW�0Dr[�y�Hz11@��t!�f"�.$�C$��&� `�Կ�3�Rbx��!R C��00D c��&�Hm1`��)�!Fl"�2��C�X�$� `�4ΐd6��!JC��0D�i(b�Hl En@��!n"�6��C�߆��`��#6��bx��!҈C��0D�qh��HiQv@��! o"�:$� C�q�$� `�T��7�Zb@��!R�C��0D*{h"�HxM�@�i�! u"�>��C�� `�t��3L�F0b|� �!�F�00D��Hg1<@�i#�l"�b$� C�[�&� `��Ѥ6��1��!�KF��0D��Ȃ�Hu1>�)4#l�"�fĈ C�܌� `�T��$7��3��!ҊF��0D�ш��HsIz@��T#�o"�j�� C�w�� `��#:�N6�H��!��F��0DjۈB�H�1<@��u#�g"�n��C��``��6L��7b��!R G�0D��H"�HgIf@�)�� m"�r4� C�_�"� `��Q�6�B:��!�MG��@0D�(�Ho1b@�i�#�g"�v4� C$ݎ&� `��e7�B<��!R�G��0D:�H��Hya~@�)�#t"�zĈC�t�&� `��d:�^>�P��!��G�0Dz��H{1>��#�g�"�~Ā C�ߏ� `��3��?��!R J��0D�AiB�Hn(Mj@�)��m"ɢ� C�q�,� `��3P�JRb� �!RLJ��0D�I��Hi)Mr@��/��n"��D� C��(� `�t��7�RT��!ҎJ��0D�S��H*1�@�iU��t"��4�C�x�(� `�T��5�vV��!��J�|0DB]��H�+Y�@��~%.>��"Űd�C�3�,� `��:@@�rY�T��!�;K��0D*i�R�H;-Y�@�)��'?�"��d�C$ۖ(< `��;�D\�d�!��K�0D�s ��H�.q��I�%�v"��C$��0� `�$��;�T_�t��!��KЪ0D� �*�H"85�@��%vY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�Bb�C�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1�p.�5��.�5��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�P3&6��3'6��3(6�p.06��.16��.26�/36�0/46�`/56��/66��/;6��/<6� 2=6�P2>6��2?6��2`6��2a6�P3d6��3e6��3f6�p.n6��.o6��.p6�/q6�0/r6�`/s6��/t6��/y6��/z6� 2{6�P2|6��2}6��2~6��26�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3 7�p.(7��.)7��.*7�/+7�0/,7�`/-7��/.7��/37��/47� 257�P267��277��287��297�P3<7��3=7��3>7�P6f7��a0�p6g7��a0��6h7��6i7��a0��6j7��6k7��a0�7l7�07q7�P7r7�p7s7��7t7��7u7��7v7��7w7�P:z7�p:{7��:|7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6 � �M#i#��ڌ`0��6$� �M #qc��ܬ`0�P7,� �M#y��޼`0��70� �M#�C��`0��:6� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� dv�\��4Ma A,$�� dv�\��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,D�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,t�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� dv�\��4Ma A,d�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,t�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,4�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� 6C�A8 v�\��`��4M��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,$�� 6C�A8 v�\��`��4M��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,T�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma Ĉ�� pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a Ĉ�� G� �pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �Q~Bp�p r2b` �~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a �� G� �� ć�B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5��`NSٽ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D��@̀�;B!8Me<B!8M�[=B!Q}�<B!8M�;!8Me��4U]�=�!8MU㶏`��ܷ &��Txo� ��4U��9B:BX:B X<!�n��,n[`�!˂�6�`NS��#@3Դ�#��T�m��4��[>!�a �� Gz �� ćzB�@v��`��#�,>Ra �� z A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&Nd� ��4��?�!�r��@̀�;B!8Me�<B!8M�<B!8M�;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}[>!�a �� ć� �� B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|ğ\p�@�@T� WC��5��`NSٽ�#�,v�`NSٶ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D�PNS�&�`NS��#��T��#@3�V�PHGT�Ǝ@NS�<B!8M�=�!8MU�f�`NSո &��T�m� ��4U��9B:BX:B X<!�n��,�}�?�!˂��`��#@3Դ�#��T6a �� ~ �� B�@v��`��#�,>Ra �� ć~ A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|ğ\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&��T�n��,˭�;4n��4�m��4Un��4o� �Y=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}[>!�a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a Ĉ�� ć� �� pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|ȟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m� �,��<B!8M��<B!8M�[=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|ԟ\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|ԟ\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me�=B!Q}[<B!8M�[;!8Me��4U]�=�!8MU�6�PNS�`NSἝ#��#��#��#��f�`��-&��Tym� ��4U��>��=&��Tyo� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|؟\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me[<B!8M��<B!8Mś=B!Q}[;!8Me��4U]�=�!8MU㶘`NS�!&��T8o�`�A`��`� ��4Uޛ<!�n� �,��c�!8M�憐0�`� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a �� G� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� Gz �p1��Q�D8��0�c�`�A`��#��T��#��T��`��a �� ć� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� ~ �p1��Q�D8��0�c�`�A`��4�n��#��T��`��a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �a~Bp�p r2b` �~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �a~Bp�p q1b` �~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v��@��a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a ��A� ��v� ��#�,a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �ǿ�v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �Ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �ǻ�v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǹ�v� ��#�,a ��A� �G��v��@��a ��A� �Ǹ�v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǯ�v��@��a ��A� �G��v��@��a ��A� �ǭ�v��@��a ��A� �G��v��@��a ��A� �Ǭ�v��@��a ��A� �G��v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǫ�v��@��a ��A� �G��v��@��a ��A� �ǩ�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a �� ȟ B�A8v��0@�c��,a ��A� �Ǟ�v��@��a ��A� �H��v��@��a ��A� �ȝ�v��@��a ��A� �G��v��@��a ��A� �ǜ�v��@��a ��A� �H��v��@��a ��A� �ț�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� Ț B�A8v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �ǘ�v��@��a ��A� �G��v��@��a ��A� �Ǐ�v��@��a ��A� �G��v��@��a ��A� �ǎ�v��@��a �� Ȏ B��@v��@H�Y:!<Sa �� H� B��@v��@H�Y:!<Sa ��A� �G��v��@��a ��A� �ǌ�v��@��a �� A�@��v�4��@��a ��A� �ǋ�v��@��a ��A� �G��v��@��a �� A�@��v�4��@��a ��A� �G��v��@��a ��A� �ǉ�v��@��a ��A� �G��v��@��a ��A� �ǈ�v��@��a ��A� �G��v��@��a ��A� ��v��@��a ��A� �G�v��@��a �� H A�@v��@NS�:!�a � �� HB�@v��@NS�:!�a �� ć~ A�@v��@��a �� ć~ A�@v��@��a "D,��A|�\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� A�@v��@��a �� z A�@v��@��a � �� {B�@v��@NS�:!�a �� { A�@v��@NS�:!�a �� z A�@v��@��a �� z A�@v��@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� y A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� A�@v��@��a �� Hn A�@v��@��a � �� nB�@v��@NS�:!�a �� n A�@v��@NS�:!�a �� Im A�@��0T�b�0!I��@��a �� l A�@��0T�b�0!I��@��a ��A#�` &�5��0T�b�0!I��04�a��,a ��A#�` &�5��0T�b�0!I��04�a��,a �� k A�@��0T�b�0!I��@��a �� Ik A�@��0T�b�0!I��@��a �� j A�@��0T�b�0!I��@��a �� Ij A�@��0T�b�0!I��@��a �� i A�@��0T�b�0!I��@��a �� Ii A�@��0T�b�0!I��@��a �� h A�@��0T�b�0!I��@��a �� Ih A�@��0T�b�0!I��@��a �� _ A�@��0T�b�0!I��@��a �� I_ A�@��0T�b�0!I��@��a �� ^ A�@��0T�b�0!I��@��a �� I^ A�@��0T�b�0!I��@��a �� ] A�@��0T�b�0!I��@��a �� I] A�@��0T�b�0!I��@��a �� \ A�@��0T�b�0!I��@��a �� ,r5A�@��0T�b�0!I��p�a��,a �� ,p5A�@��0T�b�0!I��p�a��,a �� I[ A�@��0T�b�0!I��@��a �� Z A�@��0T�b�0!I��@��a �� Z B�A8��0T�b�0!I��p�a��,a �ȈA� ��Y� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a �ȈA� �IY� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a ;J,@ �� a��pC�`0�0�� IZ%�1K�7�2F0bp p�V��#�`'j�,�,�1PAD!��@y(�p�$��A8�`��Խ�#��;B!0OM[=�!8̀�f�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#c��p�Q��,Ca#�`�#d�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8��`��Է�#�3F�`N3ؼ�#��@̳[=�!8�`��`��ص�#��9B:B�<�!�b��P�S�֎@��a 9J,@ �� Q�$� E��,�@#�`�#>��p�Q��,Ca#�`�#?�$�ǈ�� O-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[=�!8�`�֎@��a 9J,@ �� Q�$� E��,�@#�`�#9��p�Q��,Ca#�`�#:�$�ǈ�� N-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[;!�b��4��a 5J,@� ��,C #�`�1E�$� ��,�@#�`�2e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�,E�$� ��,�@#�`�-e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�''E�$� ��,�@#�`�'(e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�'"E�$� ��,�@#�`�'#e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |R�,�1bP |R�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a � �X�@v��@0�@X:!8Ma � �T�@v��@0�@X:!8Ma � �`�@v��@0�@X:!8Ma �� ; A�@��0T�b�0!I��@��a �� ; A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� ,'4A�@��0T�b�0!I��p�a��,a �� ,%4A�@��0T�b�0!I��p�a��,a �� ,#4A�@��0T�b�0!I��p�a��,a �� ,!4A�@��0T�b�0!I��p�a��,a �� B�A8��0T�b�0!I��p�a��,a �� I�B�A8��0T�b�0!I��p�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �ȈA� ��@NS�9�D�`��,a �ȈA� � ��@NS�9�D�`��,a �ȈA� �� @�@ ��0�c��4��#��T��#��T��#OD��@��a �ȈA� � �� @�@ ��0�c��4��#��T��#��T��#OD��@��a ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��d�@v�<a ��d�@v�<a ��d�@v�<a ��d�@v�<a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a A, ��\`d�`�n@��`tdld{@�;�;#��`�*��A8v�0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#��`P��A8v�pD� �a �� B�A8��`8E!��# )��@�Da ��A� ��v� a �� 0�d�2a ��A� ��v�@0��a �pCPa0bp �r6�� Ap�A8��@<1�9��a A,@ � �� B0��@<1�9��a �pCPa0bp �r6C� Ap�A8��@<1�9��a A,@� �� B0��@<1�9��a A, 4�\W ��8�C:�C9�C8��:��8́��} #��` a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұX:�!EB 4��#Y;� Q��# )a A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��pH�P a��#ұ��#�9��a A,@�&!P�܈A� Њ� �a� �A8 ��PX�0�0H�b�CD��P�2v�@0��X;� Qa A,@�&!P�܈A� Њ� �a� �A8��PX�0֎PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұ:�P��#�L��# )֎0HTa A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��@8��;B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0� 8��#ұ:�P��# �& )v�@0��;� Qa A,@�O6!P\Ď$��B$H0�� PX�0�PHT�#�@8��;B!��:�P��# )��,@C��a #A,@�&!P�\Î�A~�2b� ��6J�0�#�`�+`d��PX�0�� 8��#ұX:�!EB 4��#�,�m�CD֎0HT`�H��#�,a A,@�&!P�܈A� �G�# �`@+g��!�&��A� �� PX�0&�`��Է�#ұ:�P��#�L��# )֎0HT`��,a "A,@�&!P�\Î�A~�2b� ��6 �0�#�`�+`d��PX�0�� 8��#ұ�<�!�R߆�0T@d��@,`�H��#<!�a #A,@�O6!P\Ď�A~�L2b� ��6K$�hB��M�A8��PX�0� 8��#ұ=�!�R߆�0T@d��4� H��# )�0HT`��,a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�=�!Q��F�@8��<BHLDH�2:�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b� 1!Q��#6�pH��#>b��D>R�a FA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�PHT�#��L��#ұ��m��t,>b�H��#�Hm� 1!Q�d�CD��P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#�Hm��#$��D�D-��#Y:B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a JA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�PHT�#��P8��ܖ�`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#$��D�D-��#N�#��0T@a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��X<�HLDH�2��#ұ��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`�'O`d��PX�0��L��#��9��b��X>��D>�#��#R$�@C�="Q��Hm��t,>�#��#$��D�D-��#�<�!��a GA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ~�E�@��PX�0��L��#��9��b��D>�#��# ұ��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`�'O`d��PX�0��L��#��9��b��="Q��Hm��t,>�#��#N�#>r<BHLDH�2:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�zsy��`n v@`�dl@q�ld {p�<�A��#��`�>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ~�F�@��PX�0&��L��#��9��b��4� H��#N�#>b��ܖ��H��#>b[<BHLDH�2>"Q��Hm�CDF�pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#��L>b��D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0v��HT�#��#R$�@C�<B"��<BHLDH�2��=�D>�#>b�CDv�@0��<B"Q��X>B�D>r��>�"��>�#��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0f��HT�#��#$ұ��#4��D�D-��#Y:B1��9��b��D>�#>b��t,>b��X<�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0v��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��=�!��#,��D�D-��>"Q��>a PA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d�H��# �& )6��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#�Y;�HLDH�2�掠�L��#��T5n��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pn0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��_�O�H#�d��4UM<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pnPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-o� 1!Q˴<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��pn`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��pn`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#��T掠�L��#�Y;�HLDH�2��`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��@NSY;�HLDH�2�掠�L��#(��D�D-�n��4U�[<�!8MU󶎐�Lv�0<Sd��_�O�H#�d��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��4�m�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m=B��>�HD4�=B��>�HD4a 2A,� 8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma 2A,� 8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a �� G+A�@v�@0��:BQEa �� G+A�@��PT�Bv�@0��a �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@��PT�Bv�@0��a ��A� �G)#�`�0�v�@0��X:!�b�E,a ��A� ��(#�`�:�0�v�@0��X:!�b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a ��A� �)#�` ;�@� 0� ��@�@��@<1�9��b��,��#�Pa ��A� � )#�``;�@� 0� ��@�@��@<1�9��b�E,a��,a A, ��\U�C8��;�;�C9��A #��`�*��@v� 4a ��q� ��B�A8��0T�b�0!I��p�a��,q l2"�\�3�E<�^�3��>�`�3�I�b�486M�d�4XFX�f�5x�Z�h�5��]�j�5�6i�l�6�Fl�n�6��n�p�6�y�r�787}�t�7XG��v�8x��x�8��z�8�7��|�9�G��~�9�7��:@��:8��:X��;x0��;�@��;�0Ƀ��<��̃��=�ۃ��>8`ꃔ�?XP��?x��@��@��A�0(��B�09��D�I��D8�X��EX�_��Fxo��G��~��H�Џ��I�𨄮�J�ૄ��J��L8 ؄��Mh ߄��M��脼�N� ��N(��Nh ��N��N� ��O��OH ��O��O�0��O��PHP ��P�� P��P��PH��Q��Q��Q( ��QH(��Q�(0��R�(�(��R)@*��RH)�+��R�)@-��R�)�.��R*P8��SH*�9��S�*p;��S�*0=��S+�>��SH+0H��T�+�I��T�+0K��T,�L��TH,�N��T�,X��U�,�Y��U- [��UH-�\��U�-P^��U�-�_��V.`i��VH.�j��V�.`l��V�.�m��V/�o��W(/`y��Wh/�}��X�/`��X�/��X�/��X�/@��Y(8 ��YH8��Y�8@��Z�8��Z�8`��Z�8��Z9��[89��[x9`��[�9��[�9`ȅ��\:@ʅ��\X: ̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<`��_�<��_�=@��`8> ��a�>)��bx? 8��c�?�9��c�?P;��cH0=��cHH?��d�H�H��d�H�J��dI�L��dHI�N��d�IpX��e�IPZ��eJ0\��eHJ^��exJh��f�J0j��f�J0l��fKn��fHK�o��g�K�y��g�K�{��gL�}��gHLp��h�LP��h�L0��hM��hHM��h�MИ��i�M��iN��iHN��i�N��j�NЪ��jOЬ��jHO��j�O��k�Op��kXP��kHX0��k�XȆ��l�X�Ɇ��lY�ˆ��lHY�͆��l�Y�φ��m�Ypن��mZ@ۆ��mHZ�݆��m�Z�߆��n�Z�膮�n[Pꆱ�n([p놳�nH[�솵�nh[�톷�n�[�o�[P��o�[��o(\@��oX\p��px\� ��p�\��p�\��q�\`��q]��rH]0+��rh]�/��s�]�>��t�]X��u�]pi��v�] {��x^��y(^p��zH^P��{h^ʇ��|�^�ۇ��~�^��^��^��_��Ҁ(_��H_��h_.��_0=��_�I��_@M��_ Z��h�i��(h y��ՇHhP��hh��h��މ�h��h��̊�hp��i��ϋ(i��̌Hi��hi ��i��i0��i��i��j��(j@+��ϒHj�.��hj�=��j�J��Ô�j�M��j`Y��j�\��k�l��(k�{��Hk��hkЏ��k ��͙�k��͚�kЯ��k��l0��(l��ɝHl��hlp��l��lP��ߟ�l��l��m��(m.��Hm�:��ǣhm>��m�J��m�Z��m�i��m y��n�|��(n��Hn��hnp��nP��n��n��n0��o��(o ��Ho��ho ��op ��o ��o�)��o�-��x@=��(x`J��ĴHx@N��hx Z��x�]��xPn��x~��x��y��͹(y��Hy��hy@��y��y��ؽ�y`��y��zP��(z`��Hzp��hzp��z��z�.��zP=��z@J��{@N��({`[��H{�j��h{Py��{��{��{��{��|��(|��H|��h|0��|��|��|��|��}��(}� ��H}��h}@+��}@/��}�=��}�J��}�N��~�Z��(~�^��H~pn��h~ }��~Ћ��~��~��~�� (п��H��h��0�� 0��0 ��(��H�`(��h�/��<��H��Ȉ M��`\��k��(��y��H��}��h�p��`��؉`��(��X�p��Ȋй��(�@�]�v8threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_70_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L Q&ӄAB� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU H�b<QU�_�Ȫ"�aU 0H�R�9@\��ȫ"PP�Ew��h,*�e�EfQE�-J��(�k[�Hq��"K�`I\$�R�P袺H�"Hs�Eh �h/��t�E~Q E0 �)B#�a�I21�#R�@Jb�US�j�Ɗ�KB�X��"�a�0�dD%S�d,��"��iV�+0�8+�P@�� p��he��2��2\�VdZ�X�B�W��V�+ⲄX+�t�B��ኻ�!��p�^h�le��/:��2\X d[��W� �V�+#�x+��UC��ኄ~!��p�P�pe�b�a��2\aY2d\��B�`)W��+�C�P��+�S@�P��+�S@�~Q��+��iW�;ĺ��n ` `� �a�"��P��`�yE �J��E��E��;��Z�!��;�0�[D_ȋ��+B{t~F0��"�w�W��^_�_��,�|�рE �0*��%FA?�(A#��K��4�/��K��49�!��H�2\QDa!�(��b\�aX�+ֺ�ҰW�u�w�A⊸�<,�t��. q�^@$㊾��B\�w�Ÿb�;��R�W�b1�H�\,��`�X�+&��WX��/��B4��!�l,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �!��0;��q� �e�p�c0��`��q�a0�($r�8B1�`��`�@��\0�Q ��L.)��'��v@~��a��@;��8�` �|�q�b0�(�r�8R1�`$v�`�@�ڿ\0�U ��b��@;��r�8Z1�`d�A.�+�~5��x�` ��`�@��v@��r�c0��Y@�r ��x@�a��@;�qd9�1�`T��+�,�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0�[ �� Ɓ��@;��\�8r1�`TƂ+�.��Yp��` �Ph�`�@�zq��v@�.��q�b0�h�W0�_ �� @;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]��H��t�@��/U"�1� C$Qf� U"�3� C��f�Z `��LW��j��!�n3\-0D�o�R��!�3dP�@�Y��!Ҫ3�|�0D�ufZ�H��X��)��U"�<�� C��g�U `�D�̵*@<�v��!��3�0D��H.�@�@��)>"�aI�D��d��!R-6�0Db�f��H��X�@��&KV"�e�� C��l0Y `��%+�䳹d��!R�6��0D�� Hj�8�@� O�V"�j�� C$sm.[ `�ԯͤ+�`��t��!��6��0D��H��`�@��|LW"�os� C$n.] `��+�@��x��!�-7�0Dr��H%�L�@� ��IW"�uc� C��n,] `�t� �+�޻�t��!҉7��0D��Hu�,�@� ԛ�W"�z��C$[op `��2.D�ؽY �H��,�@�I��\"�|�C$�o"v `�4�M�.�H�y��!��7O�0D��H��,��)�e\�"��@�n�Y|��!�:Ѿ0D�C'��H��H!@��d"��C$|t � `��2�|�Y^��!RZ:K�0D�K'��H��D7@��6Ig"�� C$�t_ `��.D�X�a��!�:̸0D"S��"�� C$>u"s `�ԪN�.�|Ց��!�:һ0DY��H2�@�@�Il� _"ݭÀC��u�q@`��0D�]��H��H�@�)�_"��3�C$v&� `��Nt2�DىT��!�,;W�0DRg�y�H��01@�I��f"Ѵ#�C$�v&� `��+�J�a|��!Rl;�0D�m�"�H��0�@�� `��+�Z܉��!��;Q�0DzsGb�H��H�@�i�H^"��C�yw*| `��.H�h�a� �!�;@��W"�#�C$�w$ `��2��ߙH��!�>��0D��'R�Hd�\-@�i�e"��C�Y|$� `�T��t3��t��!�J>��0D�ɇ��Hb�0�@��-\�"��ÌC$�|��!�l>��0D��'2�H��D�@� D�]"��#�C$2} y `�D�T/�H��!R�>� 0D*ׇ$�H��0C�~}_ `��/�v��!R�>S�0D�g"�H��D'@��|�He"��s�C$~�� `��3��h��!�?��0DB�g��H��0�@��W"��ÀC�~p `��3.D�n�a �Hc�0�@�魟�\"��C��~$v `�4ޏ�.��!R�?P�0D��Hj�0��f\�"��@�\�a|��!��?Ҿ0Db�G��H��L!@�)柉d"��C��"� `�d��2�z�a^��!R�?L�0D��G��H'I7@�)�Ig"�!�� C$<�_ `�D�.D�`b��!/B̸0DBG��"�#�� C$��"s `��P�.��!]Bһ0D2M��H� A�@��<� _"�'ĀC��q@`�4�0D�Q��Hk I�@��O!�_"�*4�C$q�&� `�T�Pt2�L�T��!��BW�0Dr[�y�Hz11@��t!�f"�.$�C$��&� `�Կ�3�Rbx��!R C��00D c��&�Hm1`��)�!Fl"�2��C�X�$� `�4ΐd6��!JC��0D�i(b�Hl En@��!n"�6��C�߆��`��#6��bx��!҈C��0D�qh��HiQv@��! o"�:$� C�q�$� `�T��7�Zb@��!R�C��0D*{h"�HxM�@�i�! u"�>��C�� `�t��3L�F0b|� �!�F�00D��Hg1<@�i#�l"�b$� C�[�&� `��Ѥ6��1��!�KF��0D��Ȃ�Hu1>�)4#l�"�fĈ C�܌� `�T��$7��3��!ҊF��0D�ш��HsIz@��T#�o"�j�� C�w�� `��#:�N6�H��!��F��0DjۈB�H�1<@��u#�g"�n��C��``��6L��7b��!R G�0D��H"�HgIf@�)�� m"�r4� C�_�"� `��Q�6�B:��!�MG��@0D�(�Ho1b@�i�#�g"�v4� C$ݎ&� `��e7�B<��!R�G��0D:�H��Hya~@�)�#t"�zĈC�t�&� `��d:�^>�P��!��G�0Dz��H{1>��#�g�"�~Ā C�ߏ� `��3��?��!R J��0D�AiB�Hn(Mj@�)��m"ɢ� C�q�,� `��3P�JRb� �!RLJ��0D�I��Hi)Mr@��/��n"��D� C��(� `�t��7�RT��!ҎJ��0D�S��H*1�@�iU��t"��4�C�x�(� `�T��5�vV��!��J�|0DB]��H�+Y�@��~%.>��"Űd�C�3�,� `��:@@�rY�T��!�;K��0D*i�R�H;-Y�@�)��'?�"��d�C$ۖ(< `��;�D\�d�!��K�0D�s ��H�.q��I�%�v"��C$��0� `�$��;�T_�t��!��KЪ0D� �*�H"85�@��%vY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�Bb�C�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1�p.�5��.�5��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�P3&6��3'6��3(6�p.06��.16��.26�/36�0/46�`/56��/66��/;6��/<6� 2=6�P2>6��2?6��2`6��2a6�P3d6��3e6��3f6�p.n6��.o6��.p6�/q6�0/r6�`/s6��/t6��/y6��/z6� 2{6�P2|6��2}6��2~6��26�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3 7�p.(7��.)7��.*7�/+7�0/,7�`/-7��/.7��/37��/47� 257�P267��277��287��297�P3<7��3=7��3>7�P6f7��a0�p6g7��a0��6h7��6i7��a0��6j7��6k7��a0�7l7�07q7�P7r7�p7s7��7t7��7u7��7v7��7w7�P:z7�p:{7��:|7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6 � �M#i#��ڌ`0��6$� �M #qc��ܬ`0�P7,� �M#y��޼`0��70� �M#�C��`0��:6� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� dv�\��4Ma A,$�� dv�\��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,D�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,t�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� dv�\��4Ma A,d�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,t�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,4�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� 6C�A8 v�\��`��4M��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,$�� 6C�A8 v�\��`��4M��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,T�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma Ĉ�� pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a Ĉ�� G� �pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �Q~Bp�p r2b` �~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a �� G� �� ć�B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5��`NSٽ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D��@̀�;B!8Me<B!8M�[=B!Q}�<B!8M�;!8Me��4U]�=�!8MU㶏`��ܷ &��Txo� ��4U��9B:BX:B X<!�n��,n[`�!˂�6�`NS��#@3Դ�#��T�m��4��[>!�a �� Gz �� ćzB�@v��`��#�,>Ra �� z A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&Nd� ��4��?�!�r��@̀�;B!8Me�<B!8M�<B!8M�;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}[>!�a �� ć� �� B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|ğ\p�@�@T� WC��5��`NSٽ�#�,v�`NSٶ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D�PNS�&�`NS��#��T��#@3�V�PHGT�Ǝ@NS�<B!8M�=�!8MU�f�`NSո &��T�m� ��4U��9B:BX:B X<!�n��,�}�?�!˂��`��#@3Դ�#��T6a �� ~ �� B�@v��`��#�,>Ra �� ć~ A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|ğ\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&��T�n��,˭�;4n��4�m��4Un��4o� �Y=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}[>!�a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a Ĉ�� ć� �� pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|ȟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m� �,��<B!8M��<B!8M�[=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|ԟ\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|ԟ\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me�=B!Q}[<B!8M�[;!8Me��4U]�=�!8MU�6�PNS�`NSἝ#��#��#��#��f�`��-&��Tym� ��4U��>��=&��Tyo� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|؟\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me[<B!8M��<B!8Mś=B!Q}[;!8Me��4U]�=�!8MU㶘`NS�!&��T8o�`�A`��`� ��4Uޛ<!�n� �,��c�!8M�憐0�`� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a �� G� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� Gz �p1��Q�D8��0�c�`�A`��#��T��#��T��`��a �� ć� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� ~ �p1��Q�D8��0�c�`�A`��4�n��#��T��`��a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �a~Bp�p r2b` �~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �a~Bp�p q1b` �~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v��@��a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a ��A� ��v� ��#�,a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �ǿ�v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �Ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �ǻ�v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǹ�v� ��#�,a ��A� �G��v��@��a ��A� �Ǹ�v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǯ�v��@��a ��A� �G��v��@��a ��A� �ǭ�v��@��a ��A� �G��v��@��a ��A� �Ǭ�v��@��a ��A� �G��v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǫ�v��@��a ��A� �G��v��@��a ��A� �ǩ�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a �� ȟ B�A8v��0@�c��,a ��A� �Ǟ�v��@��a ��A� �H��v��@��a ��A� �ȝ�v��@��a ��A� �G��v��@��a ��A� �ǜ�v��@��a ��A� �H��v��@��a ��A� �ț�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� Ț B�A8v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �ǘ�v��@��a ��A� �G��v��@��a ��A� �Ǐ�v��@��a ��A� �G��v��@��a ��A� �ǎ�v��@��a �� Ȏ B��@v��@H�Y:!<Sa �� H� B��@v��@H�Y:!<Sa ��A� �G��v��@��a ��A� �ǌ�v��@��a �� A�@��v�4��@��a ��A� �ǋ�v��@��a ��A� �G��v��@��a �� A�@��v�4��@��a ��A� �G��v��@��a ��A� �ǉ�v��@��a ��A� �G��v��@��a ��A� �ǈ�v��@��a ��A� �G��v��@��a ��A� ��v��@��a ��A� �G�v��@��a �� H A�@v��@NS�:!�a � �� HB�@v��@NS�:!�a �� ć~ A�@v��@��a �� ć~ A�@v��@��a "D,��A|�\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� A�@v��@��a �� z A�@v��@��a � �� {B�@v��@NS�:!�a �� { A�@v��@NS�:!�a �� z A�@v��@��a �� z A�@v��@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� y A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� A�@v��@��a �� Hn A�@v��@��a � �� nB�@v��@NS�:!�a �� n A�@v��@NS�:!�a �� Im A�@��0T�b�0!I��@��a �� l A�@��0T�b�0!I��@��a ��A#�` &�5��0T�b�0!I��04�a��,a ��A#�` &�5��0T�b�0!I��04�a��,a �� k A�@��0T�b�0!I��@��a �� Ik A�@��0T�b�0!I��@��a �� j A�@��0T�b�0!I��@��a �� Ij A�@��0T�b�0!I��@��a �� i A�@��0T�b�0!I��@��a �� Ii A�@��0T�b�0!I��@��a �� h A�@��0T�b�0!I��@��a �� Ih A�@��0T�b�0!I��@��a �� _ A�@��0T�b�0!I��@��a �� I_ A�@��0T�b�0!I��@��a �� ^ A�@��0T�b�0!I��@��a �� I^ A�@��0T�b�0!I��@��a �� ] A�@��0T�b�0!I��@��a �� I] A�@��0T�b�0!I��@��a �� \ A�@��0T�b�0!I��@��a �� ,r5A�@��0T�b�0!I��p�a��,a �� ,p5A�@��0T�b�0!I��p�a��,a �� I[ A�@��0T�b�0!I��@��a �� Z A�@��0T�b�0!I��@��a �� Z B�A8��0T�b�0!I��p�a��,a �ȈA� ��Y� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a �ȈA� �IY� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a ;J,@ �� a��pC�`0�0�� IZ%�1K�7�2F0bp p�V��#�`'j�,�,�1PAD!��@y(�p�$��A8�`��Խ�#��;B!0OM[=�!8̀�f�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#c��p�Q��,Ca#�`�#d�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8��`��Է�#�3F�`N3ؼ�#��@̳[=�!8�`��`��ص�#��9B:B�<�!�b��P�S�֎@��a 9J,@ �� Q�$� E��,�@#�`�#>��p�Q��,Ca#�`�#?�$�ǈ�� O-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[=�!8�`�֎@��a 9J,@ �� Q�$� E��,�@#�`�#9��p�Q��,Ca#�`�#:�$�ǈ�� N-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[;!�b��4��a 5J,@� ��,C #�`�1E�$� ��,�@#�`�2e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�,E�$� ��,�@#�`�-e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�''E�$� ��,�@#�`�'(e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�'"E�$� ��,�@#�`�'#e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |R�,�1bP |R�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a � �X�@v��@0�@X:!8Ma � �T�@v��@0�@X:!8Ma � �`�@v��@0�@X:!8Ma �� ; A�@��0T�b�0!I��@��a �� ; A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� ,'4A�@��0T�b�0!I��p�a��,a �� ,%4A�@��0T�b�0!I��p�a��,a �� ,#4A�@��0T�b�0!I��p�a��,a �� ,!4A�@��0T�b�0!I��p�a��,a �� B�A8��0T�b�0!I��p�a��,a �� I�B�A8��0T�b�0!I��p�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �ȈA� ��@NS�9�D�`��,a �ȈA� � ��@NS�9�D�`��,a �ȈA� �� @�@ ��0�c��4��#��T��#��T��#OD��@��a �ȈA� � �� @�@ ��0�c��4��#��T��#��T��#OD��@��a ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��d�@v�<a ��d�@v�<a ��d�@v�<a ��d�@v�<a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a A, ��\`d�`�n@��`tdld{@�;�;#��`�*��A8v�0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#��`P��A8v�pD� �a �� B�A8��`8E!��# )��@�Da ��A� ��v� a �� 0�d�2a ��A� ��v�@0��a �pCPa0bp �r6�� Ap�A8��@<1�9��a A,@ � �� B0��@<1�9��a �pCPa0bp �r6C� Ap�A8��@<1�9��a A,@� �� B0��@<1�9��a A, 4�\W ��8�C:�C9�C8��:��8́��} #��` a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұX:�!EB 4��#Y;� Q��# )a A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��pH�P a��#ұ��#�9��a A,@�&!P�܈A� Њ� �a� �A8 ��PX�0�0H�b�CD��P�2v�@0��X;� Qa A,@�&!P�܈A� Њ� �a� �A8��PX�0֎PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұ:�P��#�L��# )֎0HTa A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��@8��;B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0� 8��#ұ:�P��# �& )v�@0��;� Qa A,@�O6!P\Ď$��B$H0�� PX�0�PHT�#�@8��;B!��:�P��# )��,@C��a #A,@�&!P�\Î�A~�2b� ��6J�0�#�`�+`d��PX�0�� 8��#ұX:�!EB 4��#�,�m�CD֎0HT`�H��#�,a A,@�&!P�܈A� �G�# �`@+g��!�&��A� �� PX�0&�`��Է�#ұ:�P��#�L��# )֎0HT`��,a "A,@�&!P�\Î�A~�2b� ��6 �0�#�`�+`d��PX�0�� 8��#ұ�<�!�R߆�0T@d��@,`�H��#<!�a #A,@�O6!P\Ď�A~�L2b� ��6K$�hB��M�A8��PX�0� 8��#ұ=�!�R߆�0T@d��4� H��# )�0HT`��,a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�=�!Q��F�@8��<BHLDH�2:�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b� 1!Q��#6�pH��#>b��D>R�a FA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�PHT�#��L��#ұ��m��t,>b�H��#�Hm� 1!Q�d�CD��P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#�Hm��#$��D�D-��#Y:B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a JA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�PHT�#��P8��ܖ�`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#$��D�D-��#N�#��0T@a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��X<�HLDH�2��#ұ��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`�'O`d��PX�0��L��#��9��b��X>��D>�#��#R$�@C�="Q��Hm��t,>�#��#$��D�D-��#�<�!��a GA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ~�E�@��PX�0��L��#��9��b��D>�#��# ұ��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`�'O`d��PX�0��L��#��9��b��="Q��Hm��t,>�#��#N�#>r<BHLDH�2:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�zsy��`n v@`�dl@q�ld {p�<�A��#��`�>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ~�F�@��PX�0&��L��#��9��b��4� H��#N�#>b��ܖ��H��#>b[<BHLDH�2>"Q��Hm�CDF�pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#��L>b��D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0v��HT�#��#R$�@C�<B"��<BHLDH�2��=�D>�#>b�CDv�@0��<B"Q��X>B�D>r��>�"��>�#��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0f��HT�#��#$ұ��#4��D�D-��#Y:B1��9��b��D>�#>b��t,>b��X<�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0v��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��=�!��#,��D�D-��>"Q��>a PA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d�H��# �& )6��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#�Y;�HLDH�2�掠�L��#��T5n��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pn0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��_�O�H#�d��4UM<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pnPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-o� 1!Q˴<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��pn`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��pn`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#��T掠�L��#�Y;�HLDH�2��`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��@NSY;�HLDH�2�掠�L��#(��D�D-�n��4U�[<�!8MU󶎐�Lv�0<Sd��_�O�H#�d��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��4�m�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m=B��>�HD4�=B��>�HD4a 2A,� 8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma 2A,� 8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a �� G+A�@v�@0��:BQEa �� G+A�@��PT�Bv�@0��a �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@��PT�Bv�@0��a ��A� �G)#�`�0�v�@0��X:!�b�E,a ��A� ��(#�`�:�0�v�@0��X:!�b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a ��A� �)#�` ;�@� 0� ��@�@��@<1�9��b��,��#�Pa ��A� � )#�``;�@� 0� ��@�@��@<1�9��b�E,a��,a A, ��\U�C8��;�;�C9��A #��`�*��@v� 4a ��q� ��B�A8��0T�b�0!I��p�a��,q l2"�\�3�E<�^�3��>�`�3�I�b�486M�d�4XFX�f�5x�Z�h�5��]�j�5�6i�l�6�Fl�n�6��n�p�6�y�r�787}�t�7XG��v�8x��x�8��z�8�7��|�9�G��~�9�7��:@��:8��:X��;x0��;�@��;�0Ƀ��<��̃��=�ۃ��>8`ꃔ�?XP��?x��@��@��A�0(��B�09��D�I��D8�X��EX�_��Fxo��G��~��H�Џ��I�𨄮�J�ૄ��J��L8 ؄��Mh ߄��M��脼�N� ��N(��Nh ��N��N� ��O��OH ��O��O�0��O��PHP ��P�� P��P��PH��Q��Q��Q( ��QH(��Q�(0��R�(�(��R)@*��RH)�+��R�)@-��R�)�.��R*P8��SH*�9��S�*p;��S�*0=��S+�>��SH+0H��T�+�I��T�+0K��T,�L��TH,�N��T�,X��U�,�Y��U- [��UH-�\��U�-P^��U�-�_��V.`i��VH.�j��V�.`l��V�.�m��V/�o��W(/`y��Wh/�}��X�/`��X�/��X�/��X�/@��Y(8 ��YH8��Y�8@��Z�8��Z�8`��Z�8��Z9��[89��[x9`��[�9��[�9`ȅ��\:@ʅ��\X: ̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<`��_�<��_�=@��`8> ��a�>)��bx? 8��c�?�9��c�?P;��cH0=��cHH?��d�H�H��d�H�J��dI�L��dHI�N��d�IpX��e�IPZ��eJ0\��eHJ^��exJh��f�J0j��f�J0l��fKn��fHK�o��g�K�y��g�K�{��gL�}��gHLp��h�LP��h�L0��hM��hHM��h�MИ��i�M��iN��iHN��i�N��j�NЪ��jOЬ��jHO��j�O��k�Op��kXP��kHX0��k�XȆ��l�X�Ɇ��lY�ˆ��lHY�͆��l�Y�φ��m�Ypن��mZ@ۆ��mHZ�݆��m�Z�߆��n�Z�膮�n[Pꆱ�n([p놳�nH[�솵�nh[�톷�n�[�o�[P��o�[��o(\@��oX\p��px\� ��p�\��p�\��q�\`��q]��rH]0+��rh]�/��s�]�>��t�]X��u�]pi��v�] {��x^��y(^p��zH^P��{h^ʇ��|�^�ۇ��~�^��^��^��_��Ҁ(_��H_��h_.��_0=��_�I��_@M��_ Z��h�i��(h y��ՇHhP��hh��h��މ�h��h��̊�hp��i��ϋ(i��̌Hi��hi ��i��i0��i��i��j��(j@+��ϒHj�.��hj�=��j�J��Ô�j�M��j`Y��j�\��k�l��(k�{��Hk��hkЏ��k ��͙�k��͚�kЯ��k��l0��(l��ɝHl��hlp��l��lP��ߟ�l��l��m��(m.��Hm�:��ǣhm>��m�J��m�Z��m�i��m y��n�|��(n��Hn��hnp��nP��n��n��n0��o��(o ��Ho��ho ��op ��o ��o�)��o�-��x@=��(x`J��ĴHx@N��hx Z��x�]��xPn��x~��x��y��͹(y��Hy��hy@��y��y��ؽ�y`��y��zP��(z`��Hzp��hzp��z��z�.��zP=��z@J��{@N��({`[��H{�j��h{Py��{��{��{��{��|��(|��H|��h|0��|��|��|��|��}��(}� ��H}��h}@+��}@/��}�=��}�J��}�N��~�Z��(~�^��H~pn��h~ }��~Ћ��~��~��~�� (п��H��h��0�� 0��0 ��(��H�`(��h�/��<��H��Ȉ M��`\��k��(��y��H��}��h�p��`��؉`��(��X�p��Ȋй��(�@�]�v8threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_72_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!�y!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQh� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@��(�@ �C� "b��H�@( �� Cb��h�p��P@�`��n��@�@��0�@��0��P�l(��@�� !�!��A��!܁ʁ� e�� `��n��@�@��0�@��0��P�l��@�� ށ��a~��!�a��@H ��@I�`� �A&A1A� CP��@&Aq$��@� EP�2�Da&ASL� �3!�&�!h&�4A� B�L Q&ӄAB� �2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��c�)�� Q@A��F�"J�|J8��FE��i�(a��l$P@Aʠ�2�i�(a�)�" PP4(��8F�"J�|�� P@A!֠� ��k�)��ɧ��n4P@�p� (�7��&_8�GP@A)⠀ ( 9��&� �APP :(�� 9�i�(a�É�w�;(�s�)��ɧ�É�wa�;��k�)��ɗ'�l$�qG��i�(a�)�p��F�w�!�;� A��&_:�h��p�w��; 9G�"J�|J:�h��p�w�;�F�"J�|�p� G�w�qG)��;�EΑ��&��'�pD�q�w��;��r�i�(a�f�Q@ �"9F�"J�| i�a��[�)��G��F��K��&�B��F�! Mn��&i� @!8��B�'�HSD ��4��R�@�#M%L>�4��R�@�#M%L��4F��p��5�Q��S��5 (�*�*�HSD ��8�a#��Њ�+�HSD �O-Nc�H0��2�b��1�Q��/Nc�h0� 1�@�\#M%L>�8�q��(�-`�r�4E�0��8"0�J��`�r�4E�0��4�0��R�b�@��\#M%L~�8�T��"�s�4E�0��_~)�0��B�\#M%L~�8�t��_~)�1��s�4E�0��_~��R$��/��B�$M%L��4��4F��E�Q��S��5 (�34ISD ��8��8�a#�� 4ISD �O-Nc,Nc�H0��2Tc�d�A�Q��/Nc,Nc�h0� ��@�\$M%L>�8��8�q��(�6`qs�4E�0��8"0�J�`}s�4E�0��4��4�0��R�c�@��#M%L��P�B�( ��C�#M%L��P�B�H ��2�C9��9�Q��,čr�!��C9��9�Q��,�r�!��R�C9�C��x @��)PAtO��)��>E0��B��r(�2��/4P~�*��D��_"��/ Ce��ȡB��/ DE0��/DQ!��%��Ȣ2��/�Q�!�Ҩ�CqTr~!�<*�@ � ��KB�Ie0�ȤBЀJE��R�@,��r�!r��C9S!�!�H�2@DS ��C8�� @��S��E䔁�ȡ��TT��*�PRE�Z�W5e�(�S�AU H�b<QU�_�Ȫ"�aU 0H�R�9@\��ȫ"PP�Ew��h,*�e�EfQE�-J��(�k[�Hq��"K�`I\$�R�P袺H�"Hs�Eh �h/��t�E~Q E0 �)B#�a�I21�#R�@Jb�US�j�Ɗ�KB�X��"�a�0�dD%S�d,��"��iV�+0�8+�P@�� p��he��2��2\�VdZ�X�B�W��V�+ⲄX+�t�B��ኻ�!��p�^h�le��/:��2\X d[��W� �V�+#�x+��UC��ኄ~!��p�P�pe�b�a��2\aY2d\��B�`)W��+�C�P��+�S@�P��+�S@�~Q��+��iW�;ĺ��n ` `� �a�"��P��`�yE �J��E��E��;��Z�!��;�0�[D_ȋ��+B{t~F0��"�w�W��^_�_��,�|�рE �0*��%FA?�(A#��K��4�/��K��49�!��H�2\QDa!�(��b\�aX�+ֺ�ҰW�u�w�A⊸�<,�t��. q�^@$㊾��B\�w�Ÿb�;��R�W�b1�H�\,��`�X�+&��WX��/��B4��!�l,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �!��0;��q� �e�p�c0��`��q�a0�($r�8B1�`��`�@��\0�Q ��L.)��'��v@~��a��@;��8�` �|�q�b0�(�r�8R1�`$v�`�@�ڿ\0�U ��b��@;��r�8Z1�`d�A.�+�~5��x�` ��`�@��v@��r�c0��Y@�r ��x@�a��@;�qd9�1�`T��+�,�FPp��` �P, �`�@��I��v@�)��q�b0�hOW0�[ �� Ɓ��@;��\�8r1�`TƂ+�.��Yp��` �Ph�`�@�zq��v@�.��q�b0�h�W0�_ �� @;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��_+0D*Y��H4K@�@��l �0"�-�C��%"v`�d��$�З��!��S=�0Dja��H=LD�@�ɍ��0"�1!�C�X&��`�4�Ą$�왘V�!RJ�K�0D�i�� HjM<4@�)� �&"�6�C��&�� `�d�D����!҉RY0D�q�A�HhN@k@� � �-"�:�C�\'"� `��t��!��S_0D { �HqOD�@�I� �4"�>1�C��'�� `��\��f��!��m0D�â� �H�XX�@�I�7"�cq�C��,"� `��䲘��!Ri�{0D��H�Y�@��=)>"�g�C$-�� `��ƴ`��!қ�|0D�ӂ��H�Z �@�)Vg>"�j��C�u-�� `��3�J�@��!R��|0D��"��Hl[L�@��o�h>"�n1�C��-�� `��4�B�p��!��|0D��Hm\@�@��g>"�r�C�].�� `��E5�F��!�M�|0DZ납�Hw]0�@�i��>"�v��C��.�� `��ޅs��p��!R��}0D��b��H>^(�@� ͋e?"�y��C$>/�� `�tꅲ@�b�`��!R�I~0D2��H�^ �@�I�f?"�|��C��/�� `��@��H��!R��~0D��"��Hy_ �@�)�$?"�~��C$�/�� `��ű@�ĿH��!��~0D��H�_�@��?"��q�C�4�� `�$��"�r� B��!R�0D*C��Hmh�@�I��?"��a�C�P4�� `��F�@�\�@F��!�-Ĉ0D�E)�H�h(%@�� D"ѣ��C�~4�� `�D��@�P�H��!RL�0D�I#�H`i#@��, EE"��Q�C$�4�� `��@��XX��!�j�0D�MC)�H�i%@� < �?"��Q�C��4�� `�D��@��@��!��0D*Q��H-j�@�IF��?"�A�C�5�� `�T�F�@��@��!��0D�SC�H�j#@��T��?"��A�C$Y5�� `��F�@�~�@��!ҹ�0DbW#��H�j�@�i_��D"��q�C��5 `��"@��@��!R��0Dr[#��Hwk�@��o �?"��Q�C$�5�� `�4��@�r�H��!��0DB_#��H�k�@��~��?"��a�C$6� `�Æ1"@�l�@��!��0Dc��Hjl�@�ɍ �?"ٱQ�C�=6�� `�T��@�X�P��!�-�0D�e��H�l$#@�i� gD"��C��6� `��ӆ2"@��h`��!\�0D�k#��H�m�@�ɴ EF"��q�C��6 `�$�"#@��H��!�|�0D�o��H!n�@��čDF"��a�C�7 `��!#@��h��!Ҝ�0D�s#�H�n#@�iՍ�F"ѺA�C$\7� `��F1"@��HF��!һň0D�w#��H�n�@�i��?"��a�C$�7�� `��@��H��!��0D�{#��Hyo�@�� F"��C$�7 `�d��a#@��0��!��0D��H�o�@��G"��1�C�<�� `��@�d�0��!�0D��Hbx8@��?"��C�R<�� `�t��@��P��!�;�0D��c��H!y �@�)%��?"��C��<�� `��@��H��!�_�0D*ͣ��H�y <@�I7��G"��A�C��<�� `�4��@��!҈�0DB��H0z`@��FL"�聀 C$3=0 `��&��@��!��0D2��H�z <@��V�G�"��C$q= `��@��@��!��H|0D��H&{`@�ieL"��a� C��=0 `��&��@��!R��0D��H�{ `@�)udD"��C��= `�Ի�#��@t��!R��0D�߃ �H(|<b@��FL�"�� C$9>1 `��G$&�d�h��!�/��0D�� H�|Dd@�ɥ��L�"�� C$�>1 `�d��#&�J�`��!Rl��0D� �H�}<b@�i�FL�"�� C$?1 `�$��#&��`��!R��0Db�� H�~<b@��FL�"�� C��?1 `��#&�t�X��!R��0D�� H�<b@��FL�"��-��` �H�@C�D��!�"��0DC�1 �Hp�Df@�)��L�"�"� C$pD4 `��ȡ&0��)��!�>"��0D I�` �H&�`n@��'N�"�%�� C$�D8 `�d�'@��!�m" �0D�M$� �H�$<@�)>��G�"�'�� C$E `��H�#@��Ix��!R�" �0D�S��H��$<@��V��G�"�+��C�tE `�$��G'�`y��!��"�~0Dr[�� H}�<�@��u��N"�/�C��E�� `��@�v��!�#��0D:e�� H��8x@�i�O�"�3�� C�F<; `��ȃ'@��y��!�^#�0DRm�� H��pz@��O�"�8�� C�G> `��'@��!Ҫ#�0D�u�� H�<|@� ߑ�O�"�<�� C��G>? `�D�� C$�G2 `�4�C&@��y��!�&��0D��1�Hk�8f@�)��L"�bҘ C�ZL3 `�4� c&@��1y��!�?&��0DZ��1�H:�8f@� -��L"�e� C��L3 `�D��c&@��3q��!|&��0D��1�H,�8f@�IG�gT"�i C�:M,1 `��$&@�z5��!�&��0D:��H=�Hb@��m�KL"�n2� C��M,3 `�� e&@�F8��!�'��0Dz��"�H��Hd@�間�L"�s2� C��N�Q `�tӉ4*@��:�F��!h'Ө0D��H�\b@� �KL"�xr� C�6O.1 `�D��$&@�n=��!һ'��0D��H:�Lb@��LL"�~B� C��O.3 `��Ie&@�JP��!R*��0D�C�"�H��Ld@��L"գB� C��T,1 `��$&@��R��!j*��0D�M�2�H�Pf@��D��L"�R� C�8U,1 `�t��$&@�rU��!�*��0DY�2�H9�Pf@��m��L"��R� C��U,1 `��$&@�BX��!+��0DZc�2�H��Pf@��L"��R� C��V,1 `�tӊ$&@��Z��!h+��0D�m�2�H�Pf@��L"ѸR� C�4W,1 `��$&@�b]��!�+��0D�w��H1�Hb@�i�KL"�2� C��W��!�+�@��l `��˴0D ��Z"��"-@�Tq��H�LC�z\.4 `�Ē�&@��r��!^.�0Dj��H�Hb@��>�KL"��2� C�],3 `�T��d&@�Pu��!ү.��0D��H)�Hb@�ig�KL"��2� C��],3 `��e&@��w��!� /��0D��Hp�Db@�)�KL"��"� C�t^*3 `��ϋd&@�`z��!�Y/��0D��H��Db@��KL"��"� C�_*3 `��d&@��|��!R�/��0D��H�Hb@�)�KL"��2� C��_,3 `��d&@�h��!��/��0D��H5�Hb@��KL"�!3� C�Xd,3 `�t�e&@��!�L2��0D:KF)�Hq� l@�)/�T "�&3� C$�d(5�`�d�L�&@�N��!��2�0DRS&��H{�d�@��V�("�+�� C$e�8 `��v*@4�Җy��!�2ة�0D�]��H��t�@��/U"�1� C$Qf� U"�3� C��f�Z `��LW��j��!�n3\-0D�o�R��!�3dP�@�Y��!Ҫ3�|�0D�ufZ�H��X��)��U"�<�� C��g�U `�D�̵*@<�v��!��3�0D��H.�@�@��)>"�aI�D��d��!R-6�0Db�f��H��X�@��&KV"�e�� C��l0Y `��%+�䳹d��!R�6��0D�� Hj�8�@� O�V"�j�� C$sm.[ `�ԯͤ+�`��t��!��6��0D��H��`�@��|LW"�os� C$n.] `��+�@��x��!�-7�0Dr��H%�L�@� ��IW"�uc� C��n,] `�t� �+�޻�t��!҉7��0D��Hu�,�@� ԛ�W"�z��C$[op `��2.D�ؽY �H��,�@�I��\"�|�C$�o"v `�4�M�.�H�y��!��7O�0D��H��,��)�e\�"��@�n�Y|��!�:Ѿ0D�C'��H��H!@��d"��C$|t � `��2�|�Y^��!RZ:K�0D�K'��H��D7@��6Ig"�� C$�t_ `��.D�X�a��!�:̸0D"S��"�� C$>u"s `�ԪN�.�|Ց��!�:һ0DY��H2�@�@�Il� _"ݭÀC��u�q@`��0D�]��H��H�@�)�_"��3�C$v&� `��Nt2�DىT��!�,;W�0DRg�y�H��01@�I��f"Ѵ#�C$�v&� `��+�J�a|��!Rl;�0D�m�"�H��0�@�� `��+�Z܉��!��;Q�0DzsGb�H��H�@�i�H^"��C�yw*| `��.H�h�a� �!�;@��W"�#�C$�w$ `��2��ߙH��!�>��0D��'R�Hd�\-@�i�e"��C�Y|$� `�T��t3��t��!�J>��0D�ɇ��Hb�0�@��-\�"��ÌC$�|��!�l>��0D��'2�H��D�@� D�]"��#�C$2} y `�D�T/�H��!R�>� 0D*ׇ$�H��0C�~}_ `��/�v��!R�>S�0D�g"�H��D'@��|�He"��s�C$~�� `��3��h��!�?��0DB�g��H��0�@��W"��ÀC�~p `��3.D�n�a �Hc�0�@�魟�\"��C��~$v `�4ޏ�.��!R�?P�0D��Hj�0��f\�"��@�\�a|��!��?Ҿ0Db�G��H��L!@�)柉d"��C��"� `�d��2�z�a^��!R�?L�0D��G��H'I7@�)�Ig"�!�� C$<�_ `�D�.D�`b��!/B̸0DBG��"�#�� C$��"s `��P�.��!]Bһ0D2M��H� A�@��<� _"�'ĀC��q@`�4�0D�Q��Hk I�@��O!�_"�*4�C$q�&� `�T�Pt2�L�T��!��BW�0Dr[�y�Hz11@��t!�f"�.$�C$��&� `�Կ�3�Rbx��!R C��00D c��&�Hm1`��)�!Fl"�2��C�X�$� `�4ΐd6��!JC��0D�i(b�Hl En@��!n"�6��C�߆��`��#6��bx��!҈C��0D�qh��HiQv@��! o"�:$� C�q�$� `�T��7�Zb@��!R�C��0D*{h"�HxM�@�i�! u"�>��C�� `�t��3L�F0b|� �!�F�00D��Hg1<@�i#�l"�b$� C�[�&� `��Ѥ6��1��!�KF��0D��Ȃ�Hu1>�)4#l�"�fĈ C�܌� `�T��$7��3��!ҊF��0D�ш��HsIz@��T#�o"�j�� C�w�� `��#:�N6�H��!��F��0DjۈB�H�1<@��u#�g"�n��C��``��6L��7b��!R G�0D��H"�HgIf@�)�� m"�r4� C�_�"� `��Q�6�B:��!�MG��@0D�(�Ho1b@�i�#�g"�v4� C$ݎ&� `��e7�B<��!R�G��0D:�H��Hya~@�)�#t"�zĈC�t�&� `��d:�^>�P��!��G�0Dz��H{1>��#�g�"�~Ā C�ߏ� `��3��?��!R J��0D�AiB�Hn(Mj@�)��m"ɢ� C�q�,� `��3P�JRb� �!RLJ��0D�I��Hi)Mr@��/��n"��D� C��(� `�t��7�RT��!ҎJ��0D�S��H*1�@�iU��t"��4�C�x�(� `�T��5�vV��!��J�|0DB]��H�+Y�@��~%.>��"Űd�C�3�,� `��:@@�rY�T��!�;K��0D*i�R�H;-Y�@�)��'?�"��d�C$ۖ(< `��;�D\�d�!��K�0D�s ��H�.q��I�%�v"��C$��0� `�$��;�T_�t��!��KЪ0D� �*�H"85�@��%vY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�Bb�C�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��a0��a0�p�0��0�0�0��a0�`�0��0�0�0��0��a0��0��0��0�0�0��a0�`�0��0� �0��0��0�P�0��0�@"1�p#1�&1�0'1��)1�+1��x1��a0��y1��~1��1��1��1��1��1�p.�5��.�5��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�P3&6��3'6��3(6�p.06��.16��.26�/36�0/46�`/56��/66��/;6��/<6� 2=6�P2>6��2?6��2`6��2a6�P3d6��3e6��3f6�p.n6��.o6��.p6�/q6�0/r6�`/s6��/t6��/y6��/z6� 2{6�P2|6��2}6��2~6��26�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3�6�p.�6��.�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�P3�6��3�6��3 7�p.(7��.)7��.*7�/+7�0/,7�`/-7��/.7��/37��/47� 257�P267��277��287��297�P3<7��3=7��3>7�P6f7��a0�p6g7��a0��6h7��6i7��a0��6j7��6k7��a0�7l7�07q7�P7r7�p7s7��7t7��7u7��7v7��7w7�P:z7�p:{7��:|7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6�7�p6�7��6�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�P:�7�p:�7��:�7�P6 � �M#i#��ڌ`0��6$� �M #qc��ܬ`0�P7,� �M#y��޼`0��70� �M#�C��`0��:6� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,�� dv�\��4Ma A,$�� dv�\��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,D�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,t�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� dv�\��4Ma A,d�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,D�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,t�� dv�\��4Ma A,�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,4�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,4�� 6C�A8 v�\��`��4M��4Ma A,d�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� dv�\��4Ma A,�� dv�\��4Ma A,$�� 6C�A8 v�\��`��4M��4Ma A,T�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma A,T�� dv�\��4Ma A,�� 6C�A8 v�\��`��4M��4Ma A,�� 6D0�B�@ v�\��`��dƎ��4M��4M֎��4Ma A,�� A�6E0��lCP�Ǝ��4M��4M��Xv�\��`��d��4M֎��4Ma Ĉ�� pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �~Bp�p r2b` ��~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a Ĉ�� G� �pC`f0��,�0K0TD�B�@��0�c�`�A`��4��#�,a YE,@��#Ez�PF�w@�^pT�x�pr2b` �Q~Bp�p r2b` �~Bp�p2��2B0K T܃p2��2D0K@T��pqQ�*\ �f1��`NSٽ&$@#�� &ұX;� ��#$@#��#��Twm��4Ս?� 0��#�,6o��&�`NS�&��T8m� ��;4n��4�m��4Un��4o��4��#��T5n��4U��9B:BX<!�n��<wn��4�M�=�!8Me�V�@��>�!8Me�a �� G� �� ć�B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|ȟ��A|П\p�@�@T� WC��5��`NSٽ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D��@̀�;B!8Me<B!8M�[=B!Q}�<B!8M�;!8Me��4U]�=�!8MU㶏`��ܷ &��Txo� ��4U��9B:BX:B X<!�n��,n[`�!˂�6�`NS��#@3Դ�#��T�m��4��[>!�a �� Gz �� ćzB�@v��`��#�,>Ra �� z A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|ȟ�@��A|П\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&Nd� ��4��?�!�r��@̀�;B!8Me�<B!8M�<B!8M�;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��tD�m�`�A`��`��tD�&�,�m��5m��4�}[>!�a �� ć� �� B�@v��P��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p��A|��A|��\p��A|��A|ğ\p�@�@T� WC��5��`NSٽ�#�,v�`NSٶ&ұX;� ��#��T7n��4խ�?�!8Mu�0�`� ��4��a��D�PNS�&�`NS��#��T��#@3�V�PHGT�Ǝ@NS�<B!8M�=�!8MU�f�`NSո &��T�m� ��4U��9B:BX:B X<!�n��,�}�?�!˂��`��#@3Դ�#��T6a �� ~ �� B�@v��`��#�,>Ra �� ć~ A�@v��@��a ZA,@��Ez�PF�w@�^pT�x�PJ��^p�@@��A|��@��A|��\p�@@��A|��@��A|ğ\p�@�@T� WC��5Ə`NSټ&ұX;� ��#��T7m��4�m`�!8Mu�`��ػ�#@3�`�!8M��6�`NS�&��T�n��,˭�;4n��4�m��4Un��4o� �Y=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o�`�A`��`��tD�&�,�m��5m��4�}[>!�a Ĉ�� G� �� ć��pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|̟��A|ԟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a Ĉ�� ć� �� pC�f0��,�0K0TD B�@ ֎@N3;� 0��#��#��#��#�,��#�,a sE,@��#Uz�P��w@!�^p��x�P��^p��A|��A|��\p��A|��A|ȟ\p��7��`��7��#@�@T@WC��A��Ƙ0H�b��tDf��Ǝ0�`��4�m��4Un��4o��,�}[;!8Me��4U]�=�!8MU�v��`��ඥ#� &�,8o��5m��4�m�>!�b��4��>�!8Me�֏`NSݸ�#��T�n� ��<��`NSݼ!&��&��T8n� ��`�!8M��@̀�=B!Q}a�!8M��v�`NS��#��`NS�-&��T� a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� w�E�9� \p\��A|��\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� w�E�9� \p\@�@��A|��\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m� �,��<B!8M��<B!8M�[=B!Q};!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ dv��0�`��,a `A,@��Ez�PF�w@�^pT�x�PJ��^p��w��!Ȉ�� G� �w�E�9� \p\��A|ԟ\p\@�@TWC��<Ə`NSټ%&ұX;� ��#��T�m��4Ս`�!8Mu�&�`NSݼ &��Two��F�`NS�&��T�n� ��;4n��4�m��4Un��4o��tD�m��4��#��Tum��4U�b�!8M��v�`NS崝#��#��#��#��V�`��#�0X`Bý�<B4CM�=�!8Meۆ�`NSٸ�#��Tvn��4��[?!�a A ��A|�W dv��0�`��,��a `A,@��Ez�PF�w@�^pT�x�PJ��^p�@@��w��!Ĉ�� G� �w�E�9� \p\@�@��A|ԟ\p\@�@TWC��<��`NSټ%&ұX;� �&��T7m� ��4�m�`�!8Mu�F�`NSݹ�#@3�a�!8M��v�`NS�!&��T8n� ��;4n��4�m��4Un��4o��tD�m� �,��;!8Me��4U]�=�!8MU�v�`NSպ�#��T5o��4U��9B:BX:B X<!�n�0�P�p�6�P�PӶ�`NSٴ�#��T�m��,��#��T�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� G� �w�E�9� \p\��A|�\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me�=B!Q}[<B!8M�[;!8Me��4U]�=�!8MU�6�PNS�`NSἝ#��#��#��#��f�`��-&��Tym� ��4U��>��=&��Tyo� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a � ��A|�^ 7x`��@@� d Ǝ0�c�`�A`��`�0֎@N3�:!�a yE,@��#Uz�P��w@!�^p��x�P��^p��w��!Ȉ�� ć� �w�E�9� \p\��A|؟\p\�7��7��@ #@�@T�WC��H֏`NSټ5&$@#��A&ұ�;� �)&$@#��#��T�m� ��4Ս[`�!8Mu떘0�c� ��4��a�!8Mu�ƘP̓�Ǝ0�`� ��4^�a�!8M��@̀�;B!8Me[<B!8M��<B!8Mś=B!Q}[;!8Me��4U]�=�!8MU㶘`NS�!&��T8o�`�A`��`� ��4Uޛ<!�n� �,��c�!8M�憐0�`� 0�n��5m� �>�!8Meۖ�`NSٸ�#��Tvn��4��?!�a �� G� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� Gz �p1��Q�D8��0�c�`�A`��#��T��#��T��`��a �� ć� �p1��Q�D8 ��0�c�`�A`��#��T��#��T��@��a �� ~ �p1��Q�D8��0�c�`�A`��4�n��#��T��`��a bA,@��Uz�P��w@!�^p��x�pr2b` �~Bp�p r2b` ��~Bp�p r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �~Bp�p q1b` ��~Bp�p q 1� gpq0l@`A@\@7�\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a bA,@��Uz�P��w@!�^p��x�pr2b` �a~Bp�p r2b` �~Bp�p r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ &��!&ұ�;� ��#��Twm��4�}`�!8Mu��0�c� ��4��?�!8Mu㦏`��ؼ�#@3�`�!8M��F�`NS�&��T�m� �<4n��4�m��4Un��4o��4��#OD��=�!8MU�V�`NSպ&��T8o�`�A`��ÿ8&�@HG��a�!8M��v�`NSٴ�#��T�m��,��#��T6a bA,@��Uz�P��w@!�^p��x�pq1b` �a~Bp�p q1b` �~Bp�p q 1� ipq0l@`A@\@7A\@\@\��l�@<֏`NSٽ!&ұ�;� ��#��T7m� ��4�][`�!8Mu�&�`NSݺ�#�a�!8Mu�`��ظ&��Two��f�`NS�&��T8n� �<4n��4�m��4Un��4o��4��#��T5m��4Um[:��D��#��T�n��4U��9B:B�:��D��/��#��6�P�s`NSٴ�#�,��`NSٹ�#��T�a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v�\��@��a ��A� �G��v�\��@��a ��A� ��v��@��a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a ��A� ��v� ��#�,a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a ��A� ��v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �ǿ�v� ��#�,a �� A�@v��@��a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �Ǽ�v� ��#�,a ��A� �G��v� ��#�,a ��A� �ǻ�v� ��#�,a ��A� �G��v� ��#�,a �� A�@v��@��a ��A� �G��v� ��#�,a ��A� �ǹ�v� ��#�,a ��A� �G��v��@��a ��A� �Ǹ�v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǯ�v��@��a ��A� �G��v��@��a ��A� �ǭ�v��@��a ��A� �G��v��@��a ��A� �Ǭ�v��@��a ��A� �G��v��@��a �� A�@v��@��a ��A� �G��v��@��a ��A� �Ǫ�v��@��a ��A� �G��v��@��a ��A� �ǩ�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� A�@v��@��a �� A�@v��@��a �� ȟ B�A8v��0@�c��,a ��A� �Ǟ�v��@��a ��A� �H��v��@��a ��A� �ȝ�v��@��a ��A� �G��v��@��a ��A� �ǜ�v��@��a ��A� �H��v��@��a ��A� �ț�v��@��a ��A� �G��v��@��a �� A�@v��@��a �� Ț B�A8v��@��a �� A�@v��@��a �� A�@v��@��a ��A� �ǘ�v��@��a ��A� �G��v��@��a ��A� �Ǐ�v��@��a ��A� �G��v��@��a ��A� �ǎ�v��@��a �� Ȏ B��@v��@H�Y:!<Sa �� H� B��@v��@H�Y:!<Sa ��A� �G��v��@��a ��A� �ǌ�v��@��a �� A�@��v�4��@��a ��A� �ǋ�v��@��a ��A� �G��v��@��a �� A�@��v�4��@��a ��A� �G��v��@��a ��A� �ǉ�v��@��a ��A� �G��v��@��a ��A� �ǈ�v��@��a ��A� �G��v��@��a ��A� ��v��@��a ��A� �G�v��@��a �� H A�@v��@NS�:!�a � �� HB�@v��@NS�:!�a �� ć~ A�@v��@��a �� ć~ A�@v��@��a "D,��A|�\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� } A�@v��@��a �� A�@v��@��a �� z A�@v��@��a � �� {B�@v��@NS�:!�a �� { A�@v��@NS�:!�a �� z A�@v��@��a �� z A�@v��@��a "D,��A|��\`��#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa "D,��A|��\`@�#�`&�u��@�A��A8�`��Է�#��T��#��T��#��Tv��Ǝ@��;BS,�Oa �� y A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� x A�@v��@��a �� A�@v��@��a �� Hn A�@v��@��a � �� nB�@v��@NS�:!�a �� n A�@v��@NS�:!�a �� Im A�@��0T�b�0!I��@��a �� l A�@��0T�b�0!I��@��a ��A#�` &�5��0T�b�0!I��04�a��,a ��A#�` &�5��0T�b�0!I��04�a��,a �� k A�@��0T�b�0!I��@��a �� Ik A�@��0T�b�0!I��@��a �� j A�@��0T�b�0!I��@��a �� Ij A�@��0T�b�0!I��@��a �� i A�@��0T�b�0!I��@��a �� Ii A�@��0T�b�0!I��@��a �� h A�@��0T�b�0!I��@��a �� Ih A�@��0T�b�0!I��@��a �� _ A�@��0T�b�0!I��@��a �� I_ A�@��0T�b�0!I��@��a �� ^ A�@��0T�b�0!I��@��a �� I^ A�@��0T�b�0!I��@��a �� ] A�@��0T�b�0!I��@��a �� I] A�@��0T�b�0!I��@��a �� \ A�@��0T�b�0!I��@��a �� ,r5A�@��0T�b�0!I��p�a��,a �� ,p5A�@��0T�b�0!I��p�a��,a �� I[ A�@��0T�b�0!I��@��a �� Z A�@��0T�b�0!I��@��a �� Z B�A8��0T�b�0!I��p�a��,a �ȈA� ��Y� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a �ȈA� �IY� ��@�@ ��0�c��4��#��T��PNS�ƎPNSٖ�@��a ;J,@ �� a��pC�`0�0�� IZ%�1K�7�2F0bp p�V��#�`'j�,�,�1PAD!��@y(�p�$��A8�`��Խ�#��;B!0OM[=�!8̀�f�`N3��#�,7m��#�,wo��<vm��4��#��#��#��#�,a 9J,@ �� Q�$� E��,�@#�`�#c��p�Q��,Ca#�`�#d�$�ǈ�� Y-�,�1PAD!��@y(�p�$��A8��`��Է�#�3F�`N3ؼ�#��@̳[=�!8�`��`��ص�#��9B:B�<�!�b��P�S�֎@��a 9J,@ �� Q�$� E��,�@#�`�#>��p�Q��,Ca#�`�#?�$�ǈ�� O-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[=�!8�`�֎@��a 9J,@ �� Q�$� E��,�@#�`�#9��p�Q��,Ca#�`�#:�$�ǈ�� N-�,�1PAD!��@y(�p�$��A8��`��Է�#�36�@N3�;!0�n��,�];� 0��#��#��#�,�n��<un��4��[;!�b��4��a 5J,@� ��,C #�`�1E�$� ��,�@#�`�2e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�,E�$� ��,�@#�`�-e� � F��~�2Kp��~�T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�''E�$� ��,�@#�`�'(e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 5J,@� ��,C #�`�'"E�$� ��,�@#�`�'#e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`��4��#�3&�`N3Զ�#��m��,Kn��<��#�,��#�춎@��<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |R�,�1bP |R�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a 7J,@ �� S1K�7x��1K�7�2F0bP |�O�,�1bP |�O�,�1PAD!��@y(�p�$��A8��0�c�`��4��#�36�`N3Զ�#��m��,Kn��<��#�,��#��Ǝ@��X<�!�R�a � �X�@v��@0�@X:!8Ma � �T�@v��@0�@X:!8Ma � �`�@v��@0�@X:!8Ma �� ; A�@��0T�b�0!I��@��a �� ; A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� : A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 9 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� 8 A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� / A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� . A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� - A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� , A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� + A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� * A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a �� ) A�@��0T�b�0!I��@��a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a ��A#�` &�4��0T�b�0!I��04�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� ,'4A�@��0T�b�0!I��p�a��,a �� ,%4A�@��0T�b�0!I��p�a��,a �� ,#4A�@��0T�b�0!I��p�a��,a �� ,!4A�@��0T�b�0!I��p�a��,a �� B�A8��0T�b�0!I��p�a��,a �� I�B�A8��0T�b�0!I��p�a��,a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �� A�@��0T�b�0!I��@��a �ȈA� ��@NS�9�D�`��,a �ȈA� � ��@NS�9�D�`��,a �ȈA� �� @�@ ��0�c��4��#��T��#��T��#OD��@��a �ȈA� � �� @�@ ��0�c��4��#��T��#��T��#OD��@��a ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��A� ��v�0<Sa ��A� �J��v�0<Sa ��d�@v�<a ��d�@v�<a ��d�@v�<a ��d�@v�<a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a ��p�@v�pD� �a A, ��\`d�`�n@��`tdld{@�;�;#��`�*��A8v�0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#��`P��A8v�pD� �a �� B�A8��`8E!��# )��@�Da ��A� ��v� a �� 0�d�2a ��A� ��v�@0��a �pCPa0bp �r6�� Ap�A8��@<1�9��a A,@ � �� B0��@<1�9��a �pCPa0bp �r6C� Ap�A8��@<1�9��a A,@� �� B0��@<1�9��a A, 4�\W ��8�C:�C9�C8��:��8́��} #��` a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұX:�!EB 4��#Y;� Q��# )a A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��pH�P a��#ұ��#�9��a A,@�&!P�܈A� Њ� �a� �A8 ��PX�0�0H�b�CD��P�2v�@0��X;� Qa A,@�&!P�܈A� Њ� �a� �A8��PX�0֎PHT�#�PH��#��0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@��PX�0�� 8��#ұ:�P��#�L��# )֎0HTa A,@�&!P�\Î$��M��&@�@��PX�0֎PHT�#��@8��;B!��:�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0� 8��#ұ:�P��# �& )v�@0��;� Qa A,@�O6!P\Ď$��B$H0�� PX�0�PHT�#�@8��;B!��:�P��# )��,@C��a #A,@�&!P�\Î�A~�2b� ��6J�0�#�`�+`d��PX�0�� 8��#ұX:�!EB 4��#�,�m�CD֎0HT`�H��#�,a A,@�&!P�܈A� �G�# �`@+g��!�&��A� �� PX�0&�`��Է�#ұ:�P��#�L��# )֎0HT`��,a "A,@�&!P�\Î�A~�2b� ��6 �0�#�`�+`d��PX�0�� 8��#ұ�<�!�R߆�0T@d��@,`�H��#<!�a #A,@�O6!P\Ď�A~�L2b� ��6K$�hB��M�A8��PX�0� 8��#ұ=�!�R߆�0T@d��4� H��# )�0HT`��,a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#R$�@C�=�!Q��F�@8��<BHLDH�2:�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b� 1!Q��#6�pH��#>b��D>R�a FA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�PHT�#��L��#ұ��m��t,>b�H��#�Hm� 1!Q�d�CD��P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�PHT�#��P8��܆�`H��#��#(��D�D-Sm��t,>b�H��#�Hm��#$��D�D-��#Y:B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a JA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�PHT�#��P8��ܖ�`H��#��#(��D�D-Sm��t,>b�H��# �& )��`HT�#��#$��D�D-��#N�#��0T@a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��X<�HLDH�2��#ұ��>"Q��>a MA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`�'O`d��PX�0��L��#��9��b��X>��D>�#��#R$�@C�="Q��Hm��t,>�#��#$��D�D-��#�<�!��a GA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ~�E�@��PX�0��L��#��9��b��D>�#��# ұ��L��0T@d��@,`��t,>�#a LA,@��4WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#��`�>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`�'O`d��PX�0��L��#��9��b��="Q��Hm��t,>�#��#N�#>r<BHLDH�2:�P��#�L��#ұ��a MA,@�O�Osu��`n v@`@p�`�dl@r�{`�9��<��<�zsy��`n v@`�dl@q�ld {p�<�A��#��`�>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ~�F�@��PX�0&��L��#��9��b��4� H��#N�#>b��ܖ��H��#>b[<BHLDH�2>"Q��Hm�CDF�pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0��pH�P a��D>�#��p8��>b��t,>b��#0��D�D-S�#��0T@d�H��#N�#>b��t,>�#��L>b��D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7�)(Q0�#��`p>N`d��PX�0v��HT�#��#R$�@C�<B"��<BHLDH�2��=�D>�#>b�CDv�@0��<B"Q��X>B�D>r��>�"��>�#��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0&�pHT�#>b��t,>b��#0��D�D-S�#��0T@d��@,`�H��#ұ��<�HLDH�2��# �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��$��M�K0��0!P(܈A� ��D�� 0b`X ��E�@��PX�0f��HT�#��#$ұ��#4��D�D-��#Y:B1��9��b��D>�#>b��t,>b��X<�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0&�pHT�#>b��#��H��#��X<HLDH�2�>b�CD��P�2v�@0��=��D>�#6�pH��#>b� 1!Q��#v��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;zsu��`n v@`@p�`�dl@r�{`�;��<��<#��`P>�1�� B`��U��A� � �8�@0�hBӄ@�p4;b� �7S)(Q0�#��`p>N`d��PX�0v��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d��@,`�H��#$��#$N�#��#(ұ��#>b� 1!Q�T��a MA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�06�pHT�#>b��#��H��#��<HLDH�2�>b�CDv�@0��X:� a�b��=�!��#,��D�D-��>"Q��>a PA,@�O�ysy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#��`P>2�� B`H��U��A� �J�< �� `��U��A� Њ�P�"� 0b`X ��F�@��PX�0��HT�#��#$ұ��#4��D�D-��# N�#>�#��0T@d�H��# �& )6��HT�#>�#��8��>�#��H��#��#8��D�D-S�#>a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#�Y;�HLDH�2�掠�L��#��T5n��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pn0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pn0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��_�O�H#�d��4UM<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��pnPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��pnPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-o� 1!Q˴<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��pn`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��pn`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#��T掠�L��#�Y;�HLDH�2��`NSո�#��T5o� 1!Q�d��_�O�H#�d��_�o�H#�d��_�_�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��@NSY;�HLDH�2�掠�L��#(��D�D-�n��4U�[<�!8MU󶎐�Lv�0<Sd��_�O�H#�d��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0n0q``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0n0v@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,T��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#��`�>�0�� Bpp0�p�lC ��LƎ��L��#��;�!8MU�֎@NS�;B��>�HD4<B��>�HD4a KA,��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma .A,��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,d��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,$��7b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#��`�p�A8v�0<Sd��4a A,D�\ah�@np��0nPq``@p�`�d��D��#��`0.��A8v�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#��`B�@v�0<Sa A,��\ah�@np��0nPv@`@p�`�d��D��#��`�+��A8v�0<Sa 1A,��7�a�� n0�� nPx�@p�d��`�d ��G��<�;#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a KA,D�8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a .A,� 8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,D�!8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,T�)8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\ah�@np��0n`s `@p�`�d��D��#��`�2��A8v�0<Sa A,$�\ah�@np��0n`v@`@p�`�d��D��#��`3��A8v�0<Sa .A,4�58b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a AA,��A8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa .A,��M8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_�_�H#�d��4�m�=B��>�HD4a 2A,��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2��4M掠�L��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m=B��>�HD4�=B��>�HD4a 2A,� 8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2��4M��4M֎��L��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,T��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�>�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�>�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!Ǝ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��4�m�=B��>�HD4=B��>�HD4a 2A,��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`@?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,d��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��``?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`�?�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,$��7�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�?�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#��`�p�A8v�0<Sd��4a A,D�\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#��`0.��A8v�0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#��`B�@v�0<Sa A,��\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#��`�+��A8v�0<Sa 5A,��7Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#��`�?�0�� Bpp0�p�lC ��LƎ��L��#�Y;!8Me��4UM�;B��>�HD4<B��>�HD4a OA,D�8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#��`�?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!F��4MƎ��L��#(��D�D-�n� 1!Q��<�!8MU�v�0<Sd��4U��:BHLDH�2�<B��>�HD4�;!8Me��_�o�H#�d��_��H#�d��4�ma 2A,� 8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#��`B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#��`0B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#��`PB�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,D�!8�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#��`pB�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#��;�HLDH�2�֎��L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,T�)8�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#��`�2��A8v�0<Sa A,$�\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#��`3��A8v�0<Sa 2A,4�58�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#��`�B�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a EA,��A8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#��`�B�0�� B0�0� �0�p�lC@� !�6ĐA8Ǝ��L��#�Y;�HLDH�2�掠�L��#4~!>�#�DD��#4~!��#�DD��#$��D�D-��#4~!~�#�DD��#4~!��#�DDa 2A,��M8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#��` C�0�� B0�P�lC ��LƎ��L��#�Y;B��>�HD4�;B��>�HD4a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,T�\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,t�\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #��``C�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G #��`�C�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�C�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#��`�CC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#��` FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,T�\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#��`@C`8v�0<Sd��4��#�Pa A,�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#��`6� `8v�0<Sd�E,a A,4�\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#��` 6� `8v�0<Sd�E,a A,��\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#��`�3� `8v�0<Sd�E,a 0A,T�\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#��`@F�!��0U�B~!~��d��4n��4��#��:�TE��O\S��#�Pa DA,t�\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG#��``F�BB��@Ǝ0U�B~!��d�S,��>qM L��PNS��PNS��@NSY<�!8MU�v�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#��`�FC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P#��`�F�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#��`GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#��`:� `8v�0<Sd�E,a A,4�\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#��` :� `8v�0<Sd�E,a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#��` GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P#��`@G�AB��@֎0U�B~!��d��3E��PT�B��0U�B~!>��d�S,��>qM LƎ0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#��``GC!v�0<Sd�S,��>qM L��0U�B~!~��d�E,a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a A,D�\ix�� n s `@p�`@q�`@r�`@s�H��#��`01�� #�c��t �a A,D�\j|��0rdldld ld0{��<��<��<��<#��`01�� #�c��t �a �� G+A�@v�@0��:BQEa �� G+A�@��PT�Bv�@0��a �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@v�@0��:BQEa �� *A�@��PT�Bv�@0��a ��A� �G)#�`�0�v�@0��X:!�b�E,a ��A� ��(#�`�:�0�v�@0��X:!�b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� Ȏ(�hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a �� hB�&�İQ�C8��@<1�9��b�E,a ��A� �)#�` ;�@� 0� ��@�@��@<1�9��b��,��#�Pa ��A� � )#�``;�@� 0� ��@�@��@<1�9��b�E,a��,a A, ��\U�C8��;�;�C9��A #��`�*��@v� 4a ��q� ��B�A8��0T�b�0!I��p�a��,q l2"�\�3�E<�^�3��>�`�3�I�b�486M�d�4XFX�f�5x�Z�h�5��]�j�5�6i�l�6�Fl�n�6��n�p�6�y�r�787}�t�7XG��v�8x��x�8��z�8�7��|�9�G��~�9�7��:@��:8��:X��;x0��;�@��;�0Ƀ��<��̃��=�ۃ��>8`ꃔ�?XP��?x��@��@��A�0(��B�09��D�I��D8�X��EX�_��Fxo��G��~��H�Џ��I�𨄮�J�ૄ��J��L8 ؄��Mh ߄��M��脼�N� ��N(��Nh ��N��N� ��O��OH ��O��O�0��O��PHP ��P�� P��P��PH��Q��Q��Q( ��QH(��Q�(0��R�(�(��R)@*��RH)�+��R�)@-��R�)�.��R*P8��SH*�9��S�*p;��S�*0=��S+�>��SH+0H��T�+�I��T�+0K��T,�L��TH,�N��T�,X��U�,�Y��U- [��UH-�\��U�-P^��U�-�_��V.`i��VH.�j��V�.`l��V�.�m��V/�o��W(/`y��Wh/�}��X�/`��X�/��X�/��X�/@��Y(8 ��YH8��Y�8@��Z�8��Z�8`��Z�8��Z9��[89��[x9`��[�9��[�9`ȅ��\:@ʅ��\X: ̅��\�:΅��\�:�υ��];�م��]X;�ۅ��]�;�݅��]�;�߅��^<��^X<`��_�<��_�=@��`8> ��a�>)��bx? 8��c�?�9��c�?P;��cH0=��cHH?��d�H�H��d�H�J��dI�L��dHI�N��d�IpX��e�IPZ��eJ0\��eHJ^��exJh��f�J0j��f�J0l��fKn��fHK�o��g�K�y��g�K�{��gL�}��gHLp��h�LP��h�L0��hM��hHM��h�MИ��i�M��iN��iHN��i�N��j�NЪ��jOЬ��jHO��j�O��k�Op��kXP��kHX0��k�XȆ��l�X�Ɇ��lY�ˆ��lHY�͆��l�Y�φ��m�Ypن��mZ@ۆ��mHZ�݆��m�Z�߆��n�Z�膮�n[Pꆱ�n([p놳�nH[�솵�nh[�톷�n�[�o�[P��o�[��o(\@��oX\p��px\� ��p�\��p�\��q�\`��q]��rH]0+��rh]�/��s�]�>��t�]X��u�]pi��v�] {��x^��y(^p��zH^P��{h^ʇ��|�^�ۇ��~�^��^��^��_��Ҁ(_��H_��h_.��_0=��_�I��_@M��_ Z��h�i��(h y��ՇHhP��hh��h��މ�h��h��̊�hp��i��ϋ(i��̌Hi��hi ��i��i0��i��i��j��(j@+��ϒHj�.��hj�=��j�J��Ô�j�M��j`Y��j�\��k�l��(k�{��Hk��hkЏ��k ��͙�k��͚�kЯ��k��l0��(l��ɝHl��hlp��l��lP��ߟ�l��l��m��(m.��Hm�:��ǣhm>��m�J��m�Z��m�i��m y��n�|��(n��Hn��hnp��nP��n��n��n0��o��(o ��Ho��ho ��op ��o ��o�)��o�-��x@=��(x`J��ĴHx@N��hx Z��x�]��xPn��x~��x��y��͹(y��Hy��hy@��y��y��ؽ�y`��y��zP��(z`��Hzp��hzp��z��z�.��zP=��z@J��{@N��({`[��H{�j��h{Py��{��{��{��{��|��(|��H|��h|0��|��|��|��|��}��(}� ��H}��h}@+��}@/��}�=��}�J��}�N��~�Z��(~�^��H~pn��h~ }��~Ћ��~��~��~�� (п��H��h��0�� 0��0 ��(��H�`(��h�/��<��H��Ȉ M��`\��k��(��y��H��}��h�p��`��؉`��(��X�p��Ȋй��(�@�]�v8threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_75_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!|z!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQl� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@�� :�� ށ��a~��!�a��@�l X6�,�$� �b��X�@0 �f� �3b��x�@@ �� e ��H5��=�C;�A8��;�=�=��;��<�=�C8��<�;��;�C9��5�� !�!��A��!܁ʁ� c@��H�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9� �`a��HI�`� �A&�1ad�$�@�0�2af$K3�ƙP��<�@ah�DɄ��H�iB@M�h�T�`� �"�gB`M*�� 2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��2TP�� 5"PA1� 1��8F�"J�|�� Ȍk�)��ɧ��jPP4(�4��&_8��F�a ( 6��&� ��@ep� �c�)�� @" (��F�"J�|J8��FPP:(��i�(a��pDP@�� (��i�(a�)�"P@�Ѓ (��#��&_:�h�Q�G��8G�"J�|J:�h�Q�G��F�"J�|�p��FB}��>��BΑ��&��'�l$��G�B�i�(a�É� }��G!飏>h�s�)��ɧ�É� }��G!飏>ȑk�)��ɗ'�pD��G}��>��i�(a�)�p� GD}��G)&飏>� Jn��&i��c�)��ɧ�f�Q@2�4�E�"J�|��l$@8��i�(a�)��l$@>�i�(a�f��B�F)D�r�4E�0�H3�(�)��1�Q��SH3�(�*��1�Q��/Nc�(0��@�\#M%L>�8�Q��"�b�r�4E�0��60��`�r�4E�0��4��(�-@�#M%L��4ƍ0��`��5�Q��S��70��B�b�P/�HSD ��8��#�0��`�0�HSD �O-Nc��0�(�0`�4��5�Q��L5 �)�1��1�HSD �_,0�H(��2$SL1��5�Q��L7�)��B0SL1�P3�HSD �_,0�(��b�)�3�SL1�A�Q��/Nc,Nc�(0��@�\$M%L>�8��8�Q��"TcYs�4E�0��60��`es�4E�0��4��4��(�6@�$M%L��4��4ƍ0��`��E�Q��S��70��B�c��8ISD ��8��8��#�0��`�P9ISD �O-Nc,Nc��0�(�9`��9�Q��,D�u��:��9�Q��,��u�)�;�C�#M%L��P�B�hPGu �:�C�#M%L��P�B��PGu�)=ꨣ�:TO��O!� h�RT@��b�"�T��PL1TP�Q�*C1�C ��b��ʠ�:PCTL�)�!��IT�b�1�(*D1��YT�!� QL1�C�bz�8*C1C�Au�!�ʠ�:�HeP�u��2`�dRPB��<T�T0�X*� `�T ( ��$S)�QGu � ��:�C6Bu��p*�@:��u ��`�:�S�1��@B�A �ʠN1�Q�N�AJJ�P@KA� ��"P=U�+��p�TF@�BTZ�U��*FU�V�(�b��*B1�W�0��*�:�$Vh��*�Y�G�@�Ң5��Z��Q�;ܢ��"�E��.��Ex�)E8E/ڋ�)6}�_�F��#=��`DR�C��h�"L�)F��ƨ1R��<F��6E�F2��<D�A��2J��dF��3Ex�3��)�;�he��H�2\QDZ�@�2�W(�V�+�Z�T+�n�A��ኸ!��p]�le��.NH�2\��)D[��W��V�+Ӄt+��B��&!��pc�pe��1UH�2\!�6D\�x�2�W �!W�+2K��+��=C̕��!��te�bH�2�:D]!Ⱥ2T�]!�H�2T��]!�ȻB�: x��>� ��(b�*�E@�J��W�ԡ��Z��!��[$��E��>��E��H�"ħ�g(�}�E@0 �� F�>è�"ܗX��c�`��4B1��BCSL1�AB�hB#��N*��,,�@⊵�4,�n��"�>� �q]}��Gb!��H�b\��X�+�2�2�W�b1��T,��X�+룏>� qc@0��H�B\!��"�b�!�qg�SL1�c!��LuHG,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �a��0;��{� �e�p�c0��H�`��q�b0�($r�8j1�`��`�@��\0�[ ��L..��'�#��v@~�ơ��@;��` �|�q�b0�(�r�8z1�`$v�`�@�ڿ\0�_ ��b��@;��r�8�1�`d�A.�8�~5��` ��`�@��v@��r�c0��Y@�| ��x@ȡ��@;�qd9�1�`T��+�9�FPp�8�` �P, �`�@��I��#��v@�)��q�c0�hOW0�u �� @;��\�8�1�`TƂ+�;��Yp�x�` �Ph�`�@�zq��#��v@�.��q�c0�h�W0�y �� A��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��m�@�f��q�`�4�� @�d�x��!R�P:p0D�_b�H!LD�@�I��HP"�1!�C�:&$|�`�t�Ĵ4�t��!�:�Hp0D�g"* �H,MH(@�ɧ i%P"�51�C��&��`�d�E4�ԛhn��!�}�N0D2q�� H4N@`@��̉g,"�9�C$R' � `�D넴�̝x��!R�P\0D*y��H5ODw@��H/"�=!�C��'$� `��4�ܟ�N��!��j0Dz�Bj �HaXL�@��i6"�bA�C�],�� `��E��H��~��!RN�x0Dz��*�H�Y\�@��7��="�g!�C��,&� `�$�U�̴��!��}0D��B��H�Z@�@�iW��>�"�j!�C$x-�� `�į��T�`��!��}0D��Hl[0�@� o'?"�n��C$�-�� `��з��!R�S~0Db�"��H=\L�@� �'?"�q�C$W.�� `��عp��!�>P~0DR��H8]@�@� ��)?"�uQ�C$�.�� `�4�E��`��!��~0D��H<^0�@� ��?"�y��C$R/�� `�$��X��!Һ�0D��b �H _$!@�)�%D"�|��C$�/�� `��E�@��`B��!��H�0D:�� H�_$!@��?"��C��/� `�d��"@��HB��!� H�0DrA#��H7h �@��$D"��q�C�74� `��"@�@�8F��!�)ň0DbE��H�h%@�I��D�"��q�C$s4� `��R"��@J��!�>E�0D�G#)�H(i%@��% �D"٤Q�C��4� `�4��"@��@V��!�[ʊ0D�Ki�H�i$-@�I5��D"ŦA�C��4� `�䛆Q"@��HR��!R{E�0D�OC��H�i0@�)D��D"��A�C$5 `�t��!#@�z�PV��!ҙƊ0DjS)�Huj%@�IO��D"��A�C�R5� `�ԪFQ"@�d�@J��!��E�0D�U�(�H�j%@�I] �D"ɫQ�C�{5� `��FQ"@�H�PR��!��F�0D�Y�(�H;k%@��g �D"��Q�C$�5� `�D�FQ"@��H��!��0DB]cY�H�k+@�Iw �F"��C$�5� `�d�FQ"@��HJ��! F�0DRa)�H2l%@�醍�D"��a�C$36� `��ƆQ"@��HJ��!�F�0D"e#)�H�l%@�i� %E"��a�C�q6� `��FQ"@��8J��!�=D�0D�g)�H#m%@� � �D"ɴq�C��6� `�4��Q"@��hR��!�]I�0Dm�I�H�m()@� � 'E"��C$�6 `��a#@�H�HJ��!R�F�0D�qC��H=n8@��̍G"��C�;7� `��Q"@�L�@J��!��E�0D�u#��H�n8@�I� G"��q�C�y7� `��R"@�T�HR��!��E�0D�y#��Hao:@��$E"��Q�C��7� `��F�"@�B�HJ��!��F�0D�}#)�H�o%@��D"��a�C$�7� `�t��Q"@��HJ��!� F�0Db��H<x$<@��G"��q�C$=<� `��P"@�F�0J��!R*C�0Db��H�x>@��D"��1�C�_<� `�T��P"@��8|��!�;��0D��C*�H$y8%@�I&�D"��C$�<� `�ė�Q"@�F�XJ��!�kH�0D��)�H�y,%@� >��D"�灔C�=� `�Ģ�Q"@�d�h��!ҏ��0D:ӣ!�Htz�@� O�>"��C$T=3 `��Ga&@�j�@��!R��0D�1�H�z f@�i^�L"�끘 C��=3 `��b&�f�@R��!��0Dۃ �Hg{ d@��mL"��A�C��=�� `�4�r�V�@��!��0D��1�H�{f@��|�L"� C$�=3 `��b&�L�@��!��0D��I�H>| b@�ɌBL�"�� C$9>0 `��ǃ&�R�`��!R-�0D"�A �H�|<h@�韏M"�� C��>4 `��G�&��h��!�j��0D��Q �H�}<h@��M�"�� C$?4 `��ǃ&��`��!R��0DB��A �H�~<h@�i�M�"�� C$?4 `��ǃ&�r�`��!R��0D��A �H�8h@��M�"�� C��?4 `�$�ǃ&�Ba��!R"�0D�A�A �Hh�@C$<D��!)"�@��f `��&��q��!R?"�0DrI�a �H=�@n@��-�M�"�%r� C$�D 9 `�t��'8�\)��!�l"�0DZO� �H� v@�i?�N�"�(�� C�E0< `�4�H"&@��I��!R�"��0D�S$ �H��`x@��W�DL�"�+�� C�xE1 `��H"&@�F��!��"��0D"[$ �Hm�$b@��n�DL�"�-�� C��E� `�ľ�"@��yB��!� #O�0D�a��Hx�@!@� �(D�"�2�C$tF� `��'�hq��!�X#��0D�k�� H~�8~@��O"�7�� C��F? `��'@�ny��!��#�0Du� �H��8�@��GT�"�;� C$}G&�0D2��a�H5�4l@�I,�M"�e° C$�L6 `��I�&@�p3y��!�y&�0D��a�H$�8l@�IF��M"�i� C�7M6 `�ԧI�&@�T5y��!R�&�0D:��I�H��0�@�IdM"�l"� C$�M.4 `��Ʉ&@�h7��!��&�0D��B�H3�Lh@�Ɍ�M"�qB� C$XN.6 `�t�I�&@��9��!�L'��0D"��R�H{�Lj@�ɵ+U"�w"� C$�N�T `��ɔ*@��<��!��'�0Dj��B�H�\h@�IߓM"�|2� C$�O04 `�� &@�`?��!��'�0D��C�H2�Ph@��M"��R� C$YT06 `��&@�@R��!�M*��0DRKS�H��Pj@��6M"��"� C$�T.4 `�T�ʄ&@��T��!��*�0D�U�b�H�Tl@��_M"Ѭ"� C$�U.4 `�Էʄ&@�`W��!��*�0D�_�b�H1�Tl@�ɌM"�"� C$WV.4 `�T�ʄ&@��Y��!�K+�0Dk�b�Hy�Tl@�ɵM"��"� C$�V.4 `��ʄ&@��\��!��+�0DRu�b�H�Tl@��M"��"� C$�W.4 `�T�ʄ&@�P_��!��+�0D��B�H'�Lh@�I�k `��0D�["��2-@��q��H%�HC��\��!�[.�@�I4��M"��B� C��\07 `�T�K�&@�tt��!�.�0D��B�H��Lh@��]�M"��2� C��].6 `�Ķ�&@�@w��!��.�0DB��B�H��Lh@�I��M"��B� C$>^.6 `�4�K�&@��y��!R?/�0Dr��B�Hd�Hh@�ɮ��M"��"� C��^,6 `�4��&@�L|��!Ҏ/�0Db��B�H��Hh@��֗�M"��2� C$~_,6 `�$��&@��~��!�/�0Dr��B�H�Lh@� ��M"� 3� C�0d.6 `��&@�X��!�82�0D�G�B�H+�Lh@��'�M"�%C� C$�d.6 `��L�&@��QV�!�2��0D2Q�Z�H<�Lr@��M Np"�*� C$Ze&8�`�t�$'@���!�2Y��0D [&��Hr�d��iu�)"�.�� C$�e�8 `�T��*@<�z��!3]��0D�e�{�H��|�@�饙0x�Hp�x�@�ɵ��"�7��C$g�Z `�t�W�F�1�W `��C��g�� `��L�@@�ʞ�b��!�3V�0Dr}F��H��h�@�I�'V"�`s� C$4l�� `��d+@H�V��!�.6�}0DZ�f��H��DC��l&\ `�$�ͥ+@H��p��!k6�0D��H��`�@� FW"�is� C�?m.\ `�d��+�ҵq��!ҽ6��0D*��H;�`�@�in��W"�n3�C��m&p `��.�L��!7�0D��H��\�@�i��\"�s3�C��n,q `�4� &.��!�l7�0D*�f�H��X�@��\"�y��C$>o0p `�D��.�ֽ��!��7�0Dj�f!�H8�,��i��\�"�}��C��o��!�7�0Dz�b�H��@�@��H^"��C�t| `��γ/��С��!�(:��@0DrEgI(�H��,C$rtr `��N2�@҉H��!RL:��0DKGR�Hu�@-@��4f"զc�C��t�� `�d��r3�Bԉt��!��:Q�0DSG�Hu�0�@�)T�\"��ØC�\us``�T��.L��a �H��0�@�)e��]"��C��u$y `�Է�T/�^ׁ��!��:о0Dz_��H$�0�� &]�"��@��a��!;R�0D2eG"�H��L'@�i��Ie"��C��v"� `��53��aj��!�h;��0DzmG��H��H=@�i�� l"��ÐC$wr `��c.L��a�� !��;L�00Du��"��ÐC$]w"v `��N�.��ݑ��!��;R�0D{��Hp�@�@� ��_"վØC��w�t�`��0D��!�H%�H!@��d"��3�C$>|&� `��O�2��`��!R<>��0DBɇ��H4�07@� ,Ig"��#�C$�|&� `�t�C.��a��!�{>��00D�χ1&�H'�0��iF `��C.��!�>ѻ0Dj�G��H��H�@�)^_"��C��}* `�d�c.P��a� �!��>@�Iu�\"��#�C$�}$� `��t2�Z��T��!?Q�0D��'��H��\3@�)��f"��C�x~$� `�4ҏ�3�j��!Z?�0D��!�H��0��\�"��äC$�~��!|?�0D��'b�H-�D�@��I^"��#�C$Q | `�$��/��!ҽ?��@0D��I(�H/�0C��r `�t��2��H��!��?��0D��gR�H��D-@�I�f"� t�C$5�� `�d�s3�D�t��!-BR�0D2Gh�H�1�@��$!�\"�$ĘC��s``��.L��b �H� 1�@��5��]"�&�C��$y `��T/�f��!ҘBо0D�S��H� 1��V!&]�"�+�@��b��!�BR�0DRYH"�H<M'@��m�Ie"�.�C�م"� `�D��53��bj��!� C��0D�aH��HeI=@�鎡 l"�2Ĉ C$[�� `�$�C6T��b�� !�>C��P0D2i�A�H2 1b@�ɧ!Im"�5$� C$ц&� `�T��7��!R~C�0DRqȲ�H`1d��!�l�"�9Ġ C$R�� `��Є7��!��C�0DRy��H>I�@� �!It"�>��C$݇� `�d��:��T��!R F�0D��r�Hl1b@� #Fl"�bĐ C$X��`�ċc6T��1b��!=F��0D�HR�H2Il@��,��m"�e4� C$Ռ"� `��QD7��3��!��F�`0Drш1,�H:1h@��L#Fl"�i4� C�R�&� `��7��5��!�F�0D��H�Hda�@��o#�t"�nĠC$ڍ&� `�t��:��7�\��!�G��0D��Hf1d@�I�#�l�"�qĘ C$U�� `�d�#6��9��!=G�0D2�hr�H9Mp@��Hn"�u� C$׎,� `�D�C6X��;b��!G�0DB��H4Mx@��̣Io"�yD� C$W�(� `�$�:��=�D��!��G�0D��1�Hj1�@��Iu"�~4�C$ޏ(� `��#6��?b��! J�P0D�A�!*�H`(1f@�� %m"�Ĉ C$R�$� `�d��6��Q��!�>J�0DrI)��H?)Et@� .%�n"��Đ C$ٔ��`��6��Sb��!�}J�0DJQi��H<*Q|@� N%�o"��$�C$[�$� `��F:�@VbL��!�J�0D�YiR�Hk+M�@��o%�u"ɮd�C$�� `��jX��!K�}0De�� $�H�,Y�@��%v"��C$��,� `��ג;�d[�`��!zK��0D�o�� H-.Y�@�i�%�O�"ݹ��C�W�8�@`��f;�F^�t��!��K� 0D�{ ��H�/a�@��%Lw"��C��0� `��+��p��!�+N��p0D��H�8}�@�I&��>"��C$Ԝ�� `��w@��s��!R�N�}0D��H�:u�@��\��>"��C$�� `��w@�`w��!��N�}0D"��H;<}�@�� wY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B)b0E�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��b0��b0��0��0��0��b0��0�`�0��0� �0��b0�P�0��0��0��0��b0��0�P�0��0��0�@�0��0�!1��$1��%1�`(1��)1�+1�`-1� "z1��b0� "{1� "�1� "�1� "�1� "�1� "�1� "�1��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�3$6�@3%6��3(6��3)6�6*6��.26�/36�0/46�`/56��/66��/76��/86� 2=6�P2>6��2?6��2`6��2a6�3b6�@3c6��3f6��3g6�6h6��.p6�/q6�0/r6�`/s6��/t6��/u6��/v6� 2{6�P2|6��2}6��2~6��26�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3 7��3!7�6"7��.*7�/+7�0/,7�`/-7��/.7��//7��/07� 257�P267��277��287��297�3:7�@3;7��3>7��3?7�6`7��6h7��b0��6i7��b0��6j7�7k7��b0�07l7�P7m7��b0�p7n7��7s7��7t7��7u7��7v7�:w7�0:x7�P:y7��:|7��:}7��:~7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6"� ��#oC��Bܔ`0�07&� �� #w��B޴`0��7.� ��#��B��`0�0:2� ��#�c��B��`0��:8� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,T�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,D�� 6C�A8 ��\Ə`��4M��4Ma A,t�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,4�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,$�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,4�� 6C�A8 ��\Ə`��4M��4Ma A,d�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� d��\֏��4Ma A,T�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma Ĉ�� ܇� �pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` ��BpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` �qBpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a �� B�@��Ə֏��P��@��a [A,@��EzAPF�}@�^�T�xAPJ��^��}��}�` �� #�`p$�)Ȉ�� ܇ BpApqQA+\ �f5�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D&�PNS�f�`NS�&��T��`NSյ&��T�%&�շ)&��T5n� ��4o�`� �,n�`4n��`�A`� �,�}[a!�n� �,��e�!8M�߆�`NS�&@3Դ-&��T�m� ��4��[c!�a �� ~ �� B�@��Ə֏�`��#�,>Ra �� ܇~ A�@��Ə֏@��a [A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}�` ��#�`p$�)Ĉ�� ܇ BpApqQA+\ �f5�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� �,˭�`4n� ��4�m� ��4Un� ��4o� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� �,�9&��T� a �� G�B�@��Ə֏��P��@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��}��}��\��}�` � #�`p#�}�� D�p5�A85�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D6�@̀�`B!8MeaB!8M�[bB!Q}�aB!8M�`!8Me� ��4U]�b�!8MU��`��ܷY&��Txo� ��4U��>B?BX?B Xa!�n� �,n[e�!˂�v�`NS�&@3Դ-&��T�m� ��4��[c!�a �� G� �� ܇�B�@��Ə֏�`��#�,>Ra �� A�@��Ə֏@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}��\��@@��}�` �#�`p#�}�� D�p5�A85�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� ��4�m� ��4Un� ��&��T�M&�,�n� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� ��4�}[c!�a Ĉ�� pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}�` �� #�`p%�)Ȉ�� BpAp2��2B0K T�p2��2D0K@T��pqQ-\ �fA֙��0H�b� �tD��0�`� ��4�m� ��4Un� ��4o� �,�}[`!8Me� ��4U]�b�!8MU㶏ƏF�`��#�Y&�,8o� �5m� ��4�m�c!�b� ��4��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}��\��}�` � #�`p$�}��pCh0��,�0P�vH��pC@h0�0�,1P�|00�� D�p5�A8A֙��0H�b� �tD��0�`� ��4�m� �,�}�aB!8M�[aB!8M�[`!8Me� ��4U]�b�!8MU㶏Ə֏F�`��Y&�,8o� �5m� ��4�m[c�!8Me��@��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a � ��}�_ d��Ə֏�0�`��,a aA,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� ܇BpAprbB� � #�`p*�}�%� D�p5�A8<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� ��4Un� ��4o� �tD�m� ��4�!&��Tum� ��4U�g�!8M�۶�`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�W d��Ə֏�0�`��,��a aA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� ܇BpApqaB��#�`p*�}�%� D�p5�A8<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�[bB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ d��Ə֏�0�`��,a `A,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� }�E�9� \�\��}�` ��'��l�@<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� �tD�m� ��4Un� ��4�!&��Tum� ��4U��aB!8M��f�!8M��ƙ`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�X d��Ə֏�0�`��,��a `A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� }�E�9� \�\@�@��}�` ��'��l�@<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�`!8Me� ��4U][bB!Q}�b�!8MU릘`NSո1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p+�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��V� 8�&@3�6�PNS�V�PNS妘PHGT�v�PNS��@NSYb�!8MU׶�`NSո}&��Tym� ��4��>B?BX?B �a!�n� �,��h�!8M��0�`� ��4UޛeBýbB4CMc�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p(�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��F�@̀�`B!8Me[aB!8M��bB!Q}�aB!8M�[`!8Me� ��4U]�b�!8MU��`NS�q&��T8o�`�A`��`� �tD�i&�,xn� ��4U��c��&��Tyo� 0�n� �5m� �c�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� ~ �p1��Q�D8��0�c�`�A`��&��T��#�,>b� ��4a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� G� �p1��Q�D8��0�c�`�A`��&��T�&��T�`��a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}e�!8Mu�&�0�c� ��4��d�!8Mu��`��ؼ&@3�e�!8M�Ӗ�`NS�a&��T�m� ��`B!8Me[aB!8M�a4n� ��4o� ��4��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�a!�n��ÿ8v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}�d� 0�U&��T7o� ��4Սe�!8Mu��`��ؼ&@3�e�!8M�ӆ�`NS�e&��T�m� �a4n� ��4�m� ��4Un� ��4�&��T��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�?��D��/�&��v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��Ə@��a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a ��A� �� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �ȿ�� #�,a �� A�@��Ə֏@��a ��A� �Ⱦ�� #�,a ��A� �H�� #�,a ��A� �Ƚ��Ə@��a ��A� �G��Ə@��a �� A�@��Ə֏@��a �� Ј� A�@��Ə֏@��a ��A� �Ȼ��Ə@��a ��A� �H��Ə@��a ��A� �Ⱥ��Ə@��a ��A� �H��Ə@��a ��A� �ȹ��Ə@��a ��A� �H��Ə@��a ��A� �ȸ��Ə@��a �� A�@��Ə֏@��a ��A� �ȯ��Ə@��a ��A� �H��Ə@��a ��A� �Ȯ��Ə@��a ��A� �H��Ə@��a ��A� �ȭ��Ə@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏0@�c��,a ��A� �H��Ə@��a ��A� �Ȫ��Ə@��a ��A� �H��Ə@��a ��A� �ȩ��Ə@��a ��A� �H��Ə@��a ��A� �Ȩ��Ə@��a ��A� �H��Ə@��a ��A� �ȟ��Ə@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏�@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H��Ə@��a ��A� �Ȝ��Ə@��a ��A� �H��Ə@��a ��A� �ț��Ə@��a ��A� �H��Ə@��a �� H� B��@��Ə@H�Y?!<Sa �� Ț B��@��Ə@H�Y?!<Sa ��A� �ș��Ə@��a ��A� �H��Ə@��a �� A�@Ə��4֏@��a ��A� �H��Ə@��a ��A� �ȏ��Ə@��a �� A�@Ə��4֏@��a ��A� �Ȏ��Ə@��a ��A� �H��Ə@��a ��A� �ȍ��Ə@��a ��A� �H��Ə@��a ��A� �Ȍ��Ə@��a ��A� �H��Ə@��a ��A� �ȋ��Ə@��a �� ɋ A�@��Ə֏@NS�?!�a � �� ɋB�@��Ə֏@NS�?!�a �� A�@��Ə֏@��a �� A�@��Ə֏@��a "D,��}��\`��#�``&-v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� I A�@��Ə֏@��a � �� B�@��Ə֏@NS�?!�a �� A�@��Ə֏@NS�?!�a �� ܇~ A�@��Ə֏@��a �� ܇~ A�@��Ə֏@��a "D,��}ğ\`��#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� A�@��Ə֏@��a �� z A�@��Ə֏@��a � �� {B�@��Ə֏@NS�?!�a �� { A�@��Ə֏@NS�?!�a �� y A�@Ə0T�b�0!I֏@��a �� Iy A�@Ə0T�b�0!I֏@��a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a �� Ix A�@Ə0T�b�0!I֏@��a �� o A�@Ə0T�b�0!I֏@��a �� Io A�@Ə0T�b�0!I֏@��a �� n A�@Ə0T�b�0!I֏@��a �� In A�@Ə0T�b�0!I֏@��a �� m A�@Ə0T�b�0!I֏@��a �� Im A�@Ə0T�b�0!I֏@��a �� l A�@Ə0T�b�0!I֏@��a �� Il A�@Ə0T�b�0!I֏@��a �� k A�@Ə0T�b�0!I֏@��a �� Ik A�@Ə0T�b�0!I֏@��a �� j A�@Ə0T�b�0!I֏@��a �� Ij A�@Ə0T�b�0!I֏@��a �� i A�@Ə0T�b�0!I֏@��a �� Ii A�@Ə0T�b�0!I֏@��a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� _ A�@Ə0T�b�0!I֏@��a �� I_ A�@Ə0T�b�0!I֏@��a �� _ B�A8֏0T�b�0!IƏp�a��,a �ȈA� �I^� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a �ȈA� ��]� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a ;J,@ �� a��pC�`0�0�� ^%�1K�7�2F0bp v�W��#�``'|�,�,�1P�D!��@�z(�p�$��A8F�`��Խ&��`B!0OM[b�!8̀Ӧ�`N3�&�,7m��&�,wo� ��<vm� ��4��#��#��#� &�,a 9J,@ �� ЕQ�$� E��,�@#�` &u��p�Q��,Ca#�` &v�$�ǈ�� ]-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 9J,@ �� Q�$� E��,�@#�` &p��p�Q��,Ca#�` &q�$�ǈ�� I\-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3�@��a!8�`� ��<�&�,vm� ��<��Əf�`��ػ&�Թ%&��vo� ��4��a 9J,@ �� Q�$� E��,�@#�` &k��p�Q��,Ca#�` &l�$�ǈ�� [-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 5J,@� ��,C #�`"cE�$� ��,�@#�`"de� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`">E�$� ��,�@#�`"?e� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'9E�$� ��,�@#�`�':e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'4E�$� ��,�@#�`�'5e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP �"S�,�1bP �"S�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP ��R�,�1bP ��R�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a � �X�@��Ə@0�@X?!8Ma � �T�@��Ə@0�@X?!8Ma � �`�@��Ə@0�@X?!8Ma �� H A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� - A�@Ə0T�b�0!I֏@��a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a �� , A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� ,94A�@֏0T�b�0!IƏp�a��,a �� ,74A�@֏0T�b�0!IƏp�a��,a �� ,54A�@֏0T�b�0!IƏp�a��,a �� ,34A�@֏0T�b�0!IƏp�a��,a �� I B�A8֏0T�b�0!IƏp�a��,a �� B�A8֏0T�b�0!IƏp�a��,a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �ȈA� � ��Ə@NS�>�D�`��,a �ȈA� ȉ��Ə@NS�>�D�`��,a �ȈA� � �� @�@ �0�c��4��#��T�&��T��#OD֏@��a �ȈA� ȉ�� @�@ �0�c��4��#��T�&��T��#OD֏@��a ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa �b�d�@��<a �a�d�@��<a ��d�@��<a ��d�@��<a �b�p�@��pD� �a �a�p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a A, �\`d�`�n@��`tdld{@�;�;#�`�*��A8��0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#�`P��A8��pD� �a �� G�B�A8֏`8E!��# )Ə@�Da ��A� �J�� a �� Ə0�d�2a ��A� �J��@0��a �pCPa0bp ��7�� Ap�A8Ə@<1�>��a A,@ � �� ؊�B0��Ə@<1�>��a �pCPa0bp ��7C� Ap�A8Ə@<1�>��a A,@� �� ؊�B0��Ə@<1�>��a A, 4�\W ��8�C:�C9�C8��:��8́��} #�` a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұX?�!EB 4��#Y`� Q��# )a A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#֏pH�P a� �� &ұ��#�>��a A,@�&!P�܈A� � � �a� �A8 �PX�0&�0H�b�CD֏P�2��@0��X`� Qa A,@�&!P�܈A� � � �a� �A8�PX�0�PHT�#&�PH��#Ə0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұ?�P��#�L��# )�0HTa A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#6�@8��`B!��?�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0F� 8� &ұ?�P��# )&�0HT`��4� H a A,@�O6!P\Ď$��B$H0�� PX�0&�PHT�#F�@8��`B!��?�P��# )֏�,@C��a #A,@�&!P�\Î�� 2b� ��7J�0�#�`�+rd�PX�06� 8� &ұX?�!EB 4�&�,�m�CD�0HT`�H�&�,a A,@�&!P�܈A� ��# �`�+y��!�&��A� �J�� PX�0f�`��Է &ұ?�P��#�L��# )�0HT`� �,a "A,@�&!P�\Î�� 2b� ��7 �0�#�`�+rd�PX�06� 8� &ұ�a�!�R�Ə0T@d��@,`�H�&a!�a #A,@�O6!P\Ď��M2b� ��7K$�hB��M�A8�PX�0F� 8� &ұb�!�R�Ə0T@d� �D��@0��X?� a�b� �,a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H��#R$�@C�b�!Q��Ԇ�@8��aBHLDH�2?�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�&ұ��a�HLDH�2��-& �H�#��`8��a FA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�PHT�#��`H��#�&(��D�D-Sm� �t,>b�H�)&�Hm� 1!Q�d�CD֏P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H�-&�Hm� 1!Q�d� ��#Y?B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a JA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�PHT�#�P8��֘`H��#�&(��D�D-Sm� �t,>b�H��# �& )Ƙ`HT�#�%&N�#V��LƏ0T@a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a MA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#��#R$�@C�b"Q��Hm� �t,>�#�&$��D�D-��#�a�!��a GA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0V��L�&��>��b� �D>�#�-& ұ��F��LƏ0T@d��@,`� �t,>�#a LA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#�-& ��Ƙ�H��#>baBHLDH�2?�P��#�L�&ұ��a MA,@�OPQsu��`n v@`@p�`�dl@r�{`�9��<��<�{sy��`n v@`�dl@q�ld {p�<�A��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0f��L�&��>��b��4� H�%&N�#>b� ��Ƙ�HT�#>R[c"��m� 1!Q�d�CD��pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�!&N�#>b� �t,>�#F��L>b� �D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0��HT�#��#R$�@C�aB"��aBHLDH�2��b�D>�#>b�CD��@0��aB"Q��XcB�D>r��c�"��>�#V��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0��HT�#��&$ұ��&4��D�D-��#Y?B1��>��b� �D>�#>b� �t,>b��Xa�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0��HT�#��&$ұ��&4��D�D-��!& N�#>�#Ə0T@d��@,`�H�&$��5&$N�#��1&(ұ��#>b� 1!Q�T��a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a PA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0Ƙ�HT�#��!&$ұ��&4��D�D-��%& N�#>�#Ə0T@d�H��# �& )v��HT�#>�#昐8��>�#֘�H��#��&8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� �_�O�H#�d� ��4UMaB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &��T&��L��#�Y`�HLDH�2�F�`NSո&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &(��D�D-o� 1!Q˴a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6�@NSY`�HLDH�2�&��L�&(��D�D-�n� ��4U�[a�!8MU��L��0<Sd� �_�O�H#�d� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a �� /A�@ƏPT�B��@0��a �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@ƏPT�B��@0��a �� /A�@ƏPT�B��@0��a ��A� ��-#�`p�0��@0��X?!�b�E,a ��A� �I-#�`;�0��@0��X?!�b�E,a �� -�hB�&�İQ�C8֏@<1�>��b�E,a �� -�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a ��A� �.#�`�;�@� 0� ��@�@֏@<1�>��b�E,a��,a ��A� ��-#�`�;�@� 0� ��@�@ƏPT�B֏@<1�>��b��,a A, $�\U�C8��;�;�C9��A #�`+��@�� 4a �� O*B�A8֏0T�b�0!IƏp�a��,a �� ܇)A�@��@0��X?!�b�E,a �� )A�@��@0��X?!�b�E,a �� ܇(A�@��@0��X?!�b�E,a �� (A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��?BQEX?!�a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,q y2"�^�4�uJ�`�4�L�b�48�O�d�5Xf[�f�5xv^�h�5��h�j�6��k�l�6�fo�n�7�vz�p�7�|�r�78��t�8Xg��v�8xw��x�8�ǘ�z�9��|�9�g��~�:�w��:`��:8p��;X��;x��;�`Ƀ��<�p̃��<�`σ��=�ڃ��>8�郔�>X��?x��@�0��@��A��B��.��C�?��D8�O��EX_��FxPn��G�p}��H��I�@��J�`��KP��K8ʄ��MX�ބ��N��턺�N��N��OH��O��O��O��O8��Ph� ��P��P��P(@��Ph��Q�`��Q��Q(��Qh��Q��R�)��R((�*��Rh(,��R�(�-��R�(0/��R()�8��Sh)0:��S�)�;��S�)@=��S(*�>��Sh*@H��T�*�I��T�*�K��T(+ M��Th+�N��T�+ X��U�+�Y��U(,P[��Uh,�\��U�,p^��U�,�_��V(-�i��Vh-0k��V�-�l��V�-@n��V(.�o��Wh.Py��W�.�z��W�.P|��W(/0~��WH/��X�/P��X�/И��Y�/p��Y�/��Y8��YH8��Zh80��Z�8��Z�80��[�8л��[9p��[89��\X9�ȅ��\�9�ʅ��\�9�̅��\�9�΅��\8:�؅��]x:�څ��]�:p܅��]�:Pޅ��]8;0腵�^x;ꅹ�^�;셽�^�;��^8<��_x<��_=P ��`�=��aX>�(��b�>p/��c�?�>��c�?H��d�?�I��d(H�K��dhH�M��d�H`O��e�H@Y��e(I [��ehI]��e�I�^��e�I�h��f(J�j��fhJ�l��f�J�n��f�J�x��g�J�z��g(K�|��ghK`~��g�K@��h�K ��h(L��hhL��h�L��i�L��i(M��ihM`��i�M@��j�M ��j(N��jhN��j�N ��k�N@��k(O@��khO ��k�O��k�O�Ȇ��l(X�ʆ��lhX�̆��l�X�Ά��l�X`؆��m(Y@چ��mhY ܆��m�Yކ��m�Y�߆��n(Z�醤�nhZ�놨�n�Z�n�Z`�o([��oH[��oh[��o�[@��o�[��o�[��p\P ��pH\��px\� ��p�\��q�\��q�\`��r]�)��r(]p-��rh]�9��s�]>��t�]M��u�]�^��v�]�o��w^��x(^p��zH^ଇ��{h^��|�^�؇��}�^0��~�^@��^��̀_��(_��H_@��h_@-��_p<��_�K��_`X��_�[��օh�h��ֆ(h`x��ɇHh��hh��h ��҉�hP��Ŋ�h��hp��i��(i0��Hi0��hi`��i��iP��Џ�i��i��jP��(j�*��Hj�9��hj=��j@L��j�X��j`\��j�_��k k��(k {��HkP��՘hk��ș�k@��k��kp��k@��lp��(l��Hl��hl0��l��ٟ�lP��l� ��Ơ�l��m��(m@-��Hmp<��hm0I��m�L��m`Y��m0i��֦�m`x��ɧn��(n��ߨHn ��ҩhn��n��Ϊ�n��np��ޫ�n��o��ޭ(oP��ٮHo��ho@��ү�o��ڰ�o��ձ�o�(��в�o@8��x<��(x�K��Hx�X��hx�\��x�h��x@l��x�|��xp��y ��(y`��Hy��hy`��y��y`��y��y��z��(z� ��Hz��hz��z��z`.��z=��z�K��{�X��({�\��H{�i��h{y��{��{p��{p��{ ��|��(|��H| ��h| ��|��|P��|��|��}��(} ��H}P��h}+��}�9��}�=��}`L��}@Y��~P]��(~`i��H~`m��h~�|��~��~@��~0��~0��P��(��H@��h��(��H� (��h��.��=��pJ��ȈpN��舐[��j��(��y��H�0��h�0��П��ȉ��Ю��(�p��H��x��@��(��H��ȋ`�� H��]��<threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_80_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64llvm.nvvm.associate.access.property.p0i8.p0i8llvm.nvvm.memcpy.async.p3i8.p1i8__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS_Z17__reduce_add_syncjj__reduce_add_sync_unsigned_impl_Z17__reduce_min_syncjj__reduce_min_sync_unsigned_impl_Z17__reduce_max_syncjj__reduce_max_sync_unsigned_impl_Z17__reduce_add_syncji__reduce_add_sync_signed_impl_Z17__reduce_min_syncji__reduce_min_sync_signed_impl_Z17__reduce_max_syncji__reduce_max_sync_signed_impl_Z17__reduce_and_syncjj__reduce_and_sync_unsigned_impl_Z16__reduce_or_syncjj__reduce_or_sync_unsigned_impl_Z17__reduce_xor_syncjj__reduce_xor_sync_unsigned_impl7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!|z!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQl� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@�� :�� ށ��a~��!�a��@�l X6�,�$� �b��X�@0 �f� �3b��x�@@ �� e ��H5��=�C;�A8��;�=�=��;��<�=�C8��<�;��;�C9��5�� !�!��A��!܁ʁ� c@��H�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9� �`a��HI�`� �A&�1ad�$�@�0�2af$K3�ƙP��<�@ah�DɄ��H�iB@M�h�T�`� �"�gB`M*�� 2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��2TP�� 5"PA1� 1��8F�"J�|�� Ȍk�)��ɧ��jPP4(�4��&_8��F�a ( 6��&� ��@ep� �c�)�� @" (��F�"J�|J8��FPP:(��i�(a��pDP@�� (��i�(a�)�"P@�Ѓ (��#��&_:�h�Q�G��8G�"J�|J:�h�Q�G��F�"J�|�p��FB}��>��BΑ��&��'�l$��G�B�i�(a�É� }��G!飏>h�s�)��ɧ�É� }��G!飏>ȑk�)��ɗ'�pD��G}��>��i�(a�)�p� GD}��G)&飏>� Jn��&i��c�)��ɧ�f�Q@2�4�E�"J�|��l$@8��i�(a�)��l$@>�i�(a�f��B�F)D�r�4E�0�H3�(�)��1�Q��SH3�(�*��1�Q��/Nc�(0��@�\#M%L>�8�Q��"�b�r�4E�0��60��`�r�4E�0��4��(�-@�#M%L��4ƍ0��`��5�Q��S��70��B�b�P/�HSD ��8��#�0��`�0�HSD �O-Nc��0�(�0`�4��5�Q��L5 �)�1��1�HSD �_,0�H(��2$SL1��5�Q��L7�)��B0SL1�P3�HSD �_,0�(��b�)�3�SL1�A�Q��/Nc,Nc�(0��@�\$M%L>�8��8�Q��"TcYs�4E�0��60��`es�4E�0��4��4��(�6@�$M%L��4��4ƍ0��`��E�Q��S��70��B�c��8ISD ��8��8��#�0��`�P9ISD �O-Nc,Nc��0�(�9`��9�Q��,D�u��:��9�Q��,��u�)�;�C�#M%L��P�B�hPGu �:�C�#M%L��P�B��PGu�)=ꨣ�:TO��O!� h�RT@��b�"�T��PL1TP�Q�*C1�C ��b��ʠ�:PCTL�)�!��IT�b�1�(*D1��YT�!� QL1�C�bz�8*C1C�Au�!�ʠ�:�HeP�u��2`�dRPB��<T�T0�X*� `�T ( ��$S)�QGu � ��:�C6Bu��p*�@:��u ��`�:�S�1��@B�A �ʠN1�Q�N�AJJ�P@KA� ��"P=U�+��p�TF@�BTZ�U��*FU�V�(�b��*B1�W�0��*�:�$Vh��*�Y�G�@�Ң5��Z��Q�;ܢ��"�E��.��Ex�)E8E/ڋ�)6}�_�F��#=��`DR�C��h�"L�)F��ƨ1R��<F��6E�F2��<D�A��2J��dF��3Ex�3��)�;�he��H�2\QDZ�@�2�W(�V�+�Z�T+�n�A��ኸ!��p]�le��.NH�2\��)D[��W��V�+Ӄt+��B��&!��pc�pe��1UH�2\!�6D\�x�2�W �!W�+2K��+��=C̕��!��te�bH�2�:D]!Ⱥ2T�]!�H�2T��]!�ȻB�: x��>� ��(b�*�E@�J��W�ԡ��Z��!��[$��E��>��E��H�"ħ�g(�}�E@0 �� F�>è�"ܗX��c�`��4B1��BCSL1�AB�hB#��N*��,,�@⊵�4,�n��"�>� �q]}��Gb!��H�b\��X�+�2�2�W�b1��T,��X�+룏>� qc@0��H�B\!��"�b�!�qg�SL1�c!��LuHG,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �a��0;��{� �e�p�c0��H�`��q�b0�($r�8j1�`��`�@��\0�[ ��L..��'�#��v@~�ơ��@;��` �|�q�b0�(�r�8z1�`$v�`�@�ڿ\0�_ ��b��@;��r�8�1�`d�A.�8�~5��` ��`�@��v@��r�c0��Y@�| ��x@ȡ��@;�qd9�1�`T��+�9�FPp�8�` �P, �`�@��I��#��v@�)��q�c0�hOW0�u �� @;��\�8�1�`TƂ+�;��Yp�x�` �Ph�`�@�zq��#��v@�.��q�c0�h�W0�y �� A��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��m�@�f��q�`�4�� @�d�x��!R�P:p0D�_b�H!LD�@�I��HP"�1!�C�:&$|�`�t�Ĵ4�t��!�:�Hp0D�g"* �H,MH(@�ɧ i%P"�51�C��&��`�d�E4�ԛhn��!�}�N0D2q�� H4N@`@��̉g,"�9�C$R' � `�D넴�̝x��!R�P\0D*y��H5ODw@��H/"�=!�C��'$� `��4�ܟ�N��!��j0Dz�Bj �HaXL�@��i6"�bA�C�],�� `��E��H��~��!RN�x0Dz��*�H�Y\�@��7��="�g!�C��,&� `�$�U�̴��!��}0D��B��H�Z@�@�iW��>�"�j!�C$x-�� `�į��T�`��!��}0D��Hl[0�@� o'?"�n��C$�-�� `��з��!R�S~0Db�"��H=\L�@� �'?"�q�C$W.�� `��عp��!�>P~0DR��H8]@�@� ��)?"�uQ�C$�.�� `�4�E��`��!��~0D��H<^0�@� ��?"�y��C$R/�� `�$��X��!Һ�0D��b �H _$!@�)�%D"�|��C$�/�� `��E�@��`B��!��H�0D:�� H�_$!@��?"��C��/� `�d��"@��HB��!� H�0DrA#��H7h �@��$D"��q�C�74� `��"@�@�8F��!�)ň0DbE��H�h%@�I��D�"��q�C$s4� `��R"��@J��!�>E�0D�G#)�H(i%@��% �D"٤Q�C��4� `�4��"@��@V��!�[ʊ0D�Ki�H�i$-@�I5��D"ŦA�C��4� `�䛆Q"@��HR��!R{E�0D�OC��H�i0@�)D��D"��A�C$5 `�t��!#@�z�PV��!ҙƊ0DjS)�Huj%@�IO��D"��A�C�R5� `�ԪFQ"@�d�@J��!��E�0D�U�(�H�j%@�I] �D"ɫQ�C�{5� `��FQ"@�H�PR��!��F�0D�Y�(�H;k%@��g �D"��Q�C$�5� `�D�FQ"@��H��!��0DB]cY�H�k+@�Iw �F"��C$�5� `�d�FQ"@��HJ��! F�0DRa)�H2l%@�醍�D"��a�C$36� `��ƆQ"@��HJ��!�F�0D"e#)�H�l%@�i� %E"��a�C�q6� `��FQ"@��8J��!�=D�0D�g)�H#m%@� � �D"ɴq�C��6� `�4��Q"@��hR��!�]I�0Dm�I�H�m()@� � 'E"��C$�6 `��a#@�H�HJ��!R�F�0D�qC��H=n8@��̍G"��C�;7� `��Q"@�L�@J��!��E�0D�u#��H�n8@�I� G"��q�C�y7� `��R"@�T�HR��!��E�0D�y#��Hao:@��$E"��Q�C��7� `��F�"@�B�HJ��!��F�0D�}#)�H�o%@��D"��a�C$�7� `�t��Q"@��HJ��!� F�0Db��H<x$<@��G"��q�C$=<� `��P"@�F�0J��!R*C�0Db��H�x>@��D"��1�C�_<� `�T��P"@��8|��!�;��0D��C*�H$y8%@�I&�D"��C$�<� `�ė�Q"@�F�XJ��!�kH�0D��)�H�y,%@� >��D"�灔C�=� `�Ģ�Q"@�d�h��!ҏ��0D:ӣ!�Htz�@� O�>"��C$T=3 `��Ga&@�j�@��!R��0D�1�H�z f@�i^�L"�끘 C��=3 `��b&�f�@R��!��0Dۃ �Hg{ d@��mL"��A�C��=�� `�4�r�V�@��!��0D��1�H�{f@��|�L"� C$�=3 `��b&�L�@��!��0D��I�H>| b@�ɌBL�"�� C$9>0 `��ǃ&�R�`��!R-�0D"�A �H�|<h@�韏M"�� C��>4 `��G�&��h��!�j��0D��Q �H�}<h@��M�"�� C$?4 `��ǃ&��`��!R��0DB��A �H�~<h@�i�M�"�� C$?4 `��ǃ&�r�`��!R��0D��A �H�8h@��M�"�� C��?4 `�$�ǃ&�Ba��!R"�0D�A�A �Hh�@C$<D��!)"�@��f `��&��q��!R?"�0DrI�a �H=�@n@��-�M�"�%r� C$�D 9 `�t��'8�\)��!�l"�0DZO� �H� v@�i?�N�"�(�� C�E0< `�4�H"&@��I��!R�"��0D�S$ �H��`x@��W�DL�"�+�� C�xE1 `��H"&@�F��!��"��0D"[$ �Hm�$b@��n�DL�"�-�� C��E� `�ľ�"@��yB��!� #O�0D�a��Hx�@!@� �(D�"�2�C$tF� `��'�hq��!�X#��0D�k�� H~�8~@��O"�7�� C��F? `��'@�ny��!��#�0Du� �H��8�@��GT�"�;� C$}G&�0D2��a�H5�4l@�I,�M"�e° C$�L6 `��I�&@�p3y��!�y&�0D��a�H$�8l@�IF��M"�i� C�7M6 `�ԧI�&@�T5y��!R�&�0D:��I�H��0�@�IdM"�l"� C$�M.4 `��Ʉ&@�h7��!��&�0D��B�H3�Lh@�Ɍ�M"�qB� C$XN.6 `�t�I�&@��9��!�L'��0D"��R�H{�Lj@�ɵ+U"�w"� C$�N�T `��ɔ*@��<��!��'�0Dj��B�H�\h@�IߓM"�|2� C$�O04 `�� &@�`?��!��'�0D��C�H2�Ph@��M"��R� C$YT06 `��&@�@R��!�M*��0DRKS�H��Pj@��6M"��"� C$�T.4 `�T�ʄ&@��T��!��*�0D�U�b�H�Tl@��_M"Ѭ"� C$�U.4 `�Էʄ&@�`W��!��*�0D�_�b�H1�Tl@�ɌM"�"� C$WV.4 `�T�ʄ&@��Y��!�K+�0Dk�b�Hy�Tl@�ɵM"��"� C$�V.4 `��ʄ&@��\��!��+�0DRu�b�H�Tl@��M"��"� C$�W.4 `�T�ʄ&@�P_��!��+�0D��B�H'�Lh@�I�k `��0D�["��2-@��q��H%�HC��\��!�[.�@�I4��M"��B� C��\07 `�T�K�&@�tt��!�.�0D��B�H��Lh@��]�M"��2� C��].6 `�Ķ�&@�@w��!��.�0DB��B�H��Lh@�I��M"��B� C$>^.6 `�4�K�&@��y��!R?/�0Dr��B�Hd�Hh@�ɮ��M"��"� C��^,6 `�4��&@�L|��!Ҏ/�0Db��B�H��Hh@��֗�M"��2� C$~_,6 `�$��&@��~��!�/�0Dr��B�H�Lh@� ��M"� 3� C�0d.6 `��&@�X��!�82�0D�G�B�H+�Lh@��'�M"�%C� C$�d.6 `��L�&@��QV�!�2��0D2Q�Z�H<�Lr@��M Np"�*� C$Ze&8�`�t�$'@���!�2Y��0D [&��Hr�d��iu�)"�.�� C$�e�8 `�T��*@<�z��!3]��0D�e�{�H��|�@�饙0x�Hp�x�@�ɵ��"�7��C$g�Z `�t�W�F�1�W `��C��g�� `��L�@@�ʞ�b��!�3V�0Dr}F��H��h�@�I�'V"�`s� C$4l�� `��d+@H�V��!�.6�}0DZ�f��H��DC��l&\ `�$�ͥ+@H��p��!k6�0D��H��`�@� FW"�is� C�?m.\ `�d��+�ҵq��!ҽ6��0D*��H;�`�@�in��W"�n3�C��m&p `��.�L��!7�0D��H��\�@�i��\"�s3�C��n,q `�4� &.��!�l7�0D*�f�H��X�@��\"�y��C$>o0p `�D��.�ֽ��!��7�0Dj�f!�H8�,��i��\�"�}��C��o��!�7�0Dz�b�H��@�@��H^"��C�t| `��γ/��С��!�(:��@0DrEgI(�H��,C$rtr `��N2�@҉H��!RL:��0DKGR�Hu�@-@��4f"զc�C��t�� `�d��r3�Bԉt��!��:Q�0DSG�Hu�0�@�)T�\"��ØC�\us``�T��.L��a �H��0�@�)e��]"��C��u$y `�Է�T/�^ׁ��!��:о0Dz_��H$�0�� &]�"��@��a��!;R�0D2eG"�H��L'@�i��Ie"��C��v"� `��53��aj��!�h;��0DzmG��H��H=@�i�� l"��ÐC$wr `��c.L��a�� !��;L�00Du��"��ÐC$]w"v `��N�.��ݑ��!��;R�0D{��Hp�@�@� ��_"վØC��w�t�`��0D��!�H%�H!@��d"��3�C$>|&� `��O�2��`��!R<>��0DBɇ��H4�07@� ,Ig"��#�C$�|&� `�t�C.��a��!�{>��00D�χ1&�H'�0��iF `��C.��!�>ѻ0Dj�G��H��H�@�)^_"��C��}* `�d�c.P��a� �!��>@�Iu�\"��#�C$�}$� `��t2�Z��T��!?Q�0D��'��H��\3@�)��f"��C�x~$� `�4ҏ�3�j��!Z?�0D��!�H��0��\�"��äC$�~��!|?�0D��'b�H-�D�@��I^"��#�C$Q | `�$��/��!ҽ?��@0D��I(�H/�0C��r `�t��2��H��!��?��0D��gR�H��D-@�I�f"� t�C$5�� `�d�s3�D�t��!-BR�0D2Gh�H�1�@��$!�\"�$ĘC��s``��.L��b �H� 1�@��5��]"�&�C��$y `��T/�f��!ҘBо0D�S��H� 1��V!&]�"�+�@��b��!�BR�0DRYH"�H<M'@��m�Ie"�.�C�م"� `�D��53��bj��!� C��0D�aH��HeI=@�鎡 l"�2Ĉ C$[�� `�$�C6T��b�� !�>C��P0D2i�A�H2 1b@�ɧ!Im"�5$� C$ц&� `�T��7��!R~C�0DRqȲ�H`1d��!�l�"�9Ġ C$R�� `��Є7��!��C�0DRy��H>I�@� �!It"�>��C$݇� `�d��:��T��!R F�0D��r�Hl1b@� #Fl"�bĐ C$X��`�ċc6T��1b��!=F��0D�HR�H2Il@��,��m"�e4� C$Ռ"� `��QD7��3��!��F�`0Drш1,�H:1h@��L#Fl"�i4� C�R�&� `��7��5��!�F�0D��H�Hda�@��o#�t"�nĠC$ڍ&� `�t��:��7�\��!�G��0D��Hf1d@�I�#�l�"�qĘ C$U�� `�d�#6��9��!=G�0D2�hr�H9Mp@��Hn"�u� C$׎,� `�D�C6X��;b��!G�0DB��H4Mx@��̣Io"�yD� C$W�(� `�$�:��=�D��!��G�0D��1�Hj1�@��Iu"�~4�C$ޏ(� `��#6��?b��! J�P0D�A�!*�H`(1f@�� %m"�Ĉ C$R�$� `�d��6��Q��!�>J�0DrI)��H?)Et@� .%�n"��Đ C$ٔ��`��6��Sb��!�}J�0DJQi��H<*Q|@� N%�o"��$�C$[�$� `��F:�@VbL��!�J�0D�YiR�Hk+M�@��o%�u"ɮd�C$�� `��jX��!K�}0De�� $�H�,Y�@��%v"��C$��,� `��ג;�d[�`��!zK��0D�o�� H-.Y�@�i�%�O�"ݹ��C�W�8�@`��f;�F^�t��!��K� 0D�{ ��H�/a�@��%Lw"��C��0� `��+��p��!�+N��p0D��H�8}�@�I&��>"��C$Ԝ�� `��w@��s��!R�N�}0D��H�:u�@��\��>"��C$�� `��w@�`w��!��N�}0D"��H;<}�@�� wY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B)b0E�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��b0��b0��0��0��0��b0��0�`�0��0� �0��b0�P�0��0��0��0��b0��0�P�0��0��0�@�0��0�!1��$1��%1�`(1��)1�+1�`-1� "z1��b0� "{1� "�1� "�1� "�1� "�1� "�1� "�1��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�3$6�@3%6��3(6��3)6�6*6��.26�/36�0/46�`/56��/66��/76��/86� 2=6�P2>6��2?6��2`6��2a6�3b6�@3c6��3f6��3g6�6h6��.p6�/q6�0/r6�`/s6��/t6��/u6��/v6� 2{6�P2|6��2}6��2~6��26�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3 7��3!7�6"7��.*7�/+7�0/,7�`/-7��/.7��//7��/07� 257�P267��277��287��297�3:7�@3;7��3>7��3?7�6`7��6h7��b0��6i7��b0��6j7�7k7��b0�07l7�P7m7��b0�p7n7��7s7��7t7��7u7��7v7�:w7�0:x7�P:y7��:|7��:}7��:~7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6"� ��#oC��Bܔ`0�07&� �� #w��B޴`0��7.� ��#��B��`0�0:2� ��#�c��B��`0��:8� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,T�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,D�� 6C�A8 ��\Ə`��4M��4Ma A,t�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,4�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,$�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,4�� 6C�A8 ��\Ə`��4M��4Ma A,d�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� d��\֏��4Ma A,T�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma Ĉ�� ܇� �pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` ��BpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` �qBpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a �� B�@��Ə֏��P��@��a [A,@��EzAPF�}@�^�T�xAPJ��^��}��}�` �� #�`p$�)Ȉ�� ܇ BpApqQA+\ �f5�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D&�PNS�f�`NS�&��T��`NSյ&��T�%&�շ)&��T5n� ��4o�`� �,n�`4n��`�A`� �,�}[a!�n� �,��e�!8M�߆�`NS�&@3Դ-&��T�m� ��4��[c!�a �� ~ �� B�@��Ə֏�`��#�,>Ra �� ܇~ A�@��Ə֏@��a [A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}�` ��#�`p$�)Ĉ�� ܇ BpApqQA+\ �f5�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� �,˭�`4n� ��4�m� ��4Un� ��4o� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� �,�9&��T� a �� G�B�@��Ə֏��P��@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��}��}��\��}�` � #�`p#�}�� D�p5�A85�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D6�@̀�`B!8MeaB!8M�[bB!Q}�aB!8M�`!8Me� ��4U]�b�!8MU��`��ܷY&��Txo� ��4U��>B?BX?B Xa!�n� �,n[e�!˂�v�`NS�&@3Դ-&��T�m� ��4��[c!�a �� G� �� ܇�B�@��Ə֏�`��#�,>Ra �� A�@��Ə֏@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}��\��@@��}�` �#�`p#�}�� D�p5�A85�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� ��4�m� ��4Un� ��&��T�M&�,�n� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� ��4�}[c!�a Ĉ�� pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}�` �� #�`p%�)Ȉ�� BpAp2��2B0K T�p2��2D0K@T��pqQ-\ �fA֙��0H�b� �tD��0�`� ��4�m� ��4Un� ��4o� �,�}[`!8Me� ��4U]�b�!8MU㶏ƏF�`��#�Y&�,8o� �5m� ��4�m�c!�b� ��4��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}��\��}�` � #�`p$�}��pCh0��,�0P�vH��pC@h0�0�,1P�|00�� D�p5�A8A֙��0H�b� �tD��0�`� ��4�m� �,�}�aB!8M�[aB!8M�[`!8Me� ��4U]�b�!8MU㶏Ə֏F�`��Y&�,8o� �5m� ��4�m[c�!8Me��@��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a � ��}�_ d��Ə֏�0�`��,a aA,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� ܇BpAprbB� � #�`p*�}�%� D�p5�A8<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� ��4Un� ��4o� �tD�m� ��4�!&��Tum� ��4U�g�!8M�۶�`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�W d��Ə֏�0�`��,��a aA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� ܇BpApqaB��#�`p*�}�%� D�p5�A8<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�[bB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ d��Ə֏�0�`��,a `A,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� }�E�9� \�\��}�` ��'��l�@<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� �tD�m� ��4Un� ��4�!&��Tum� ��4U��aB!8M��f�!8M��ƙ`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�X d��Ə֏�0�`��,��a `A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� }�E�9� \�\@�@��}�` ��'��l�@<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�`!8Me� ��4U][bB!Q}�b�!8MU릘`NSո1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p+�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��V� 8�&@3�6�PNS�V�PNS妘PHGT�v�PNS��@NSYb�!8MU׶�`NSո}&��Tym� ��4��>B?BX?B �a!�n� �,��h�!8M��0�`� ��4UޛeBýbB4CMc�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p(�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��F�@̀�`B!8Me[aB!8M��bB!Q}�aB!8M�[`!8Me� ��4U]�b�!8MU��`NS�q&��T8o�`�A`��`� �tD�i&�,xn� ��4U��c��&��Tyo� 0�n� �5m� �c�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� ~ �p1��Q�D8��0�c�`�A`��&��T��#�,>b� ��4a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� G� �p1��Q�D8��0�c�`�A`��&��T�&��T�`��a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}e�!8Mu�&�0�c� ��4��d�!8Mu��`��ؼ&@3�e�!8M�Ӗ�`NS�a&��T�m� ��`B!8Me[aB!8M�a4n� ��4o� ��4��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�a!�n��ÿ8v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}�d� 0�U&��T7o� ��4Սe�!8Mu��`��ؼ&@3�e�!8M�ӆ�`NS�e&��T�m� �a4n� ��4�m� ��4Un� ��4�&��T��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�?��D��/�&��v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��Ə@��a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a ��A� �� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �ȿ�� #�,a �� A�@��Ə֏@��a ��A� �Ⱦ�� #�,a ��A� �H�� #�,a ��A� �Ƚ��Ə@��a ��A� �G��Ə@��a �� A�@��Ə֏@��a �� Ј� A�@��Ə֏@��a ��A� �Ȼ��Ə@��a ��A� �H��Ə@��a ��A� �Ⱥ��Ə@��a ��A� �H��Ə@��a ��A� �ȹ��Ə@��a ��A� �H��Ə@��a ��A� �ȸ��Ə@��a �� A�@��Ə֏@��a ��A� �ȯ��Ə@��a ��A� �H��Ə@��a ��A� �Ȯ��Ə@��a ��A� �H��Ə@��a ��A� �ȭ��Ə@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏0@�c��,a ��A� �H��Ə@��a ��A� �Ȫ��Ə@��a ��A� �H��Ə@��a ��A� �ȩ��Ə@��a ��A� �H��Ə@��a ��A� �Ȩ��Ə@��a ��A� �H��Ə@��a ��A� �ȟ��Ə@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏�@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H��Ə@��a ��A� �Ȝ��Ə@��a ��A� �H��Ə@��a ��A� �ț��Ə@��a ��A� �H��Ə@��a �� H� B��@��Ə@H�Y?!<Sa �� Ț B��@��Ə@H�Y?!<Sa ��A� �ș��Ə@��a ��A� �H��Ə@��a �� A�@Ə��4֏@��a ��A� �H��Ə@��a ��A� �ȏ��Ə@��a �� A�@Ə��4֏@��a ��A� �Ȏ��Ə@��a ��A� �H��Ə@��a ��A� �ȍ��Ə@��a ��A� �H��Ə@��a ��A� �Ȍ��Ə@��a ��A� �H��Ə@��a ��A� �ȋ��Ə@��a �� ɋ A�@��Ə֏@NS�?!�a � �� ɋB�@��Ə֏@NS�?!�a �� A�@��Ə֏@��a �� A�@��Ə֏@��a "D,��}��\`��#�``&-v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� I A�@��Ə֏@��a � �� B�@��Ə֏@NS�?!�a �� A�@��Ə֏@NS�?!�a �� ܇~ A�@��Ə֏@��a �� ܇~ A�@��Ə֏@��a "D,��}ğ\`��#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� A�@��Ə֏@��a �� z A�@��Ə֏@��a � �� {B�@��Ə֏@NS�?!�a �� { A�@��Ə֏@NS�?!�a �� y A�@Ə0T�b�0!I֏@��a �� Iy A�@Ə0T�b�0!I֏@��a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a �� Ix A�@Ə0T�b�0!I֏@��a �� o A�@Ə0T�b�0!I֏@��a �� Io A�@Ə0T�b�0!I֏@��a �� n A�@Ə0T�b�0!I֏@��a �� In A�@Ə0T�b�0!I֏@��a �� m A�@Ə0T�b�0!I֏@��a �� Im A�@Ə0T�b�0!I֏@��a �� l A�@Ə0T�b�0!I֏@��a �� Il A�@Ə0T�b�0!I֏@��a �� k A�@Ə0T�b�0!I֏@��a �� Ik A�@Ə0T�b�0!I֏@��a �� j A�@Ə0T�b�0!I֏@��a �� Ij A�@Ə0T�b�0!I֏@��a �� i A�@Ə0T�b�0!I֏@��a �� Ii A�@Ə0T�b�0!I֏@��a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� _ A�@Ə0T�b�0!I֏@��a �� I_ A�@Ə0T�b�0!I֏@��a �� _ B�A8֏0T�b�0!IƏp�a��,a �ȈA� �I^� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a �ȈA� ��]� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a ;J,@ �� a��pC�`0�0�� ^%�1K�7�2F0bp v�W��#�``'|�,�,�1P�D!��@�z(�p�$��A8F�`��Խ&��`B!0OM[b�!8̀Ӧ�`N3�&�,7m��&�,wo� ��<vm� ��4��#��#��#� &�,a 9J,@ �� ЕQ�$� E��,�@#�` &u��p�Q��,Ca#�` &v�$�ǈ�� ]-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 9J,@ �� Q�$� E��,�@#�` &p��p�Q��,Ca#�` &q�$�ǈ�� I\-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3�@��a!8�`� ��<�&�,vm� ��<��Əf�`��ػ&�Թ%&��vo� ��4��a 9J,@ �� Q�$� E��,�@#�` &k��p�Q��,Ca#�` &l�$�ǈ�� [-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 5J,@� ��,C #�`"cE�$� ��,�@#�`"de� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`">E�$� ��,�@#�`"?e� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'9E�$� ��,�@#�`�':e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'4E�$� ��,�@#�`�'5e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP �"S�,�1bP �"S�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP ��R�,�1bP ��R�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a � �X�@��Ə@0�@X?!8Ma � �T�@��Ə@0�@X?!8Ma � �`�@��Ə@0�@X?!8Ma �� H A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� - A�@Ə0T�b�0!I֏@��a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a �� , A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� ,94A�@֏0T�b�0!IƏp�a��,a �� ,74A�@֏0T�b�0!IƏp�a��,a �� ,54A�@֏0T�b�0!IƏp�a��,a �� ,34A�@֏0T�b�0!IƏp�a��,a �� I B�A8֏0T�b�0!IƏp�a��,a �� B�A8֏0T�b�0!IƏp�a��,a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �ȈA� � ��Ə@NS�>�D�`��,a �ȈA� ȉ��Ə@NS�>�D�`��,a �ȈA� � �� @�@ �0�c��4��#��T�&��T��#OD֏@��a �ȈA� ȉ�� @�@ �0�c��4��#��T�&��T��#OD֏@��a ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa �b�d�@��<a �a�d�@��<a ��d�@��<a ��d�@��<a �b�p�@��pD� �a �a�p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a A, �\`d�`�n@��`tdld{@�;�;#�`�*��A8��0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#�`P��A8��pD� �a �� G�B�A8֏`8E!��# )Ə@�Da ��A� �J�� a �� Ə0�d�2a ��A� �J��@0��a �pCPa0bp ��7�� Ap�A8Ə@<1�>��a A,@ � �� ؊�B0��Ə@<1�>��a �pCPa0bp ��7C� Ap�A8Ə@<1�>��a A,@� �� ؊�B0��Ə@<1�>��a A, 4�\W ��8�C:�C9�C8��:��8́��} #�` a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұX?�!EB 4��#Y`� Q��# )a A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#֏pH�P a� �� &ұ��#�>��a A,@�&!P�܈A� � � �a� �A8 �PX�0&�0H�b�CD֏P�2��@0��X`� Qa A,@�&!P�܈A� � � �a� �A8�PX�0�PHT�#&�PH��#Ə0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұ?�P��#�L��# )�0HTa A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#6�@8��`B!��?�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0F� 8� &ұ?�P��# )&�0HT`��4� H a A,@�O6!P\Ď$��B$H0�� PX�0&�PHT�#F�@8��`B!��?�P��# )֏�,@C��a #A,@�&!P�\Î�� 2b� ��7J�0�#�`�+rd�PX�06� 8� &ұX?�!EB 4�&�,�m�CD�0HT`�H�&�,a A,@�&!P�܈A� ��# �`�+y��!�&��A� �J�� PX�0f�`��Է &ұ?�P��#�L��# )�0HT`� �,a "A,@�&!P�\Î�� 2b� ��7 �0�#�`�+rd�PX�06� 8� &ұ�a�!�R�Ə0T@d��@,`�H�&a!�a #A,@�O6!P\Ď��M2b� ��7K$�hB��M�A8�PX�0F� 8� &ұb�!�R�Ə0T@d� �D��@0��X?� a�b� �,a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H��#R$�@C�b�!Q��Ԇ�@8��aBHLDH�2?�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�&ұ��a�HLDH�2��-& �H�#��`8��a FA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�PHT�#��`H��#�&(��D�D-Sm� �t,>b�H�)&�Hm� 1!Q�d�CD֏P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H�-&�Hm� 1!Q�d� ��#Y?B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a JA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�PHT�#�P8��֘`H��#�&(��D�D-Sm� �t,>b�H��# �& )Ƙ`HT�#�%&N�#V��LƏ0T@a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a MA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#��#R$�@C�b"Q��Hm� �t,>�#�&$��D�D-��#�a�!��a GA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0V��L�&��>��b� �D>�#�-& ұ��F��LƏ0T@d��@,`� �t,>�#a LA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#�-& ��Ƙ�H��#>baBHLDH�2?�P��#�L�&ұ��a MA,@�OPQsu��`n v@`@p�`�dl@r�{`�9��<��<�{sy��`n v@`�dl@q�ld {p�<�A��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0f��L�&��>��b��4� H�%&N�#>b� ��Ƙ�HT�#>R[c"��m� 1!Q�d�CD��pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�!&N�#>b� �t,>�#F��L>b� �D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0��HT�#��#R$�@C�aB"��aBHLDH�2��b�D>�#>b�CD��@0��aB"Q��XcB�D>r��c�"��>�#V��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0��HT�#��&$ұ��&4��D�D-��#Y?B1��>��b� �D>�#>b� �t,>b��Xa�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0��HT�#��&$ұ��&4��D�D-��!& N�#>�#Ə0T@d��@,`�H�&$��5&$N�#��1&(ұ��#>b� 1!Q�T��a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a PA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0Ƙ�HT�#��!&$ұ��&4��D�D-��%& N�#>�#Ə0T@d�H��# �& )v��HT�#>�#昐8��>�#֘�H��#��&8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� �_�O�H#�d� ��4UMaB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &��T&��L��#�Y`�HLDH�2�F�`NSո&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &(��D�D-o� 1!Q˴a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6�@NSY`�HLDH�2�&��L�&(��D�D-�n� ��4U�[a�!8MU��L��0<Sd� �_�O�H#�d� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a �� /A�@ƏPT�B��@0��a �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@ƏPT�B��@0��a �� /A�@ƏPT�B��@0��a ��A� ��-#�`p�0��@0��X?!�b�E,a ��A� �I-#�`;�0��@0��X?!�b�E,a �� -�hB�&�İQ�C8֏@<1�>��b�E,a �� -�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a ��A� �.#�`�;�@� 0� ��@�@֏@<1�>��b�E,a��,a ��A� ��-#�`�;�@� 0� ��@�@ƏPT�B֏@<1�>��b��,a A, $�\U�C8��;�;�C9��A #�`+��@�� 4a �� O*B�A8֏0T�b�0!IƏp�a��,a �� ܇)A�@��@0��X?!�b�E,a �� )A�@��@0��X?!�b�E,a �� ܇(A�@��@0��X?!�b�E,a �� (A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��?BQEX?!�a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,q y2"�^�4�uJ�`�4�L�b�48�O�d�5Xf[�f�5xv^�h�5��h�j�6��k�l�6�fo�n�7�vz�p�7�|�r�78��t�8Xg��v�8xw��x�8�ǘ�z�9��|�9�g��~�:�w��:`��:8p��;X��;x��;�`Ƀ��<�p̃��<�`σ��=�ڃ��>8�郔�>X��?x��@�0��@��A��B��.��C�?��D8�O��EX_��FxPn��G�p}��H��I�@��J�`��KP��K8ʄ��MX�ބ��N��턺�N��N��OH��O��O��O��O8��Ph� ��P��P��P(@��Ph��Q�`��Q��Q(��Qh��Q��R�)��R((�*��Rh(,��R�(�-��R�(0/��R()�8��Sh)0:��S�)�;��S�)@=��S(*�>��Sh*@H��T�*�I��T�*�K��T(+ M��Th+�N��T�+ X��U�+�Y��U(,P[��Uh,�\��U�,p^��U�,�_��V(-�i��Vh-0k��V�-�l��V�-@n��V(.�o��Wh.Py��W�.�z��W�.P|��W(/0~��WH/��X�/P��X�/И��Y�/p��Y�/��Y8��YH8��Zh80��Z�8��Z�80��[�8л��[9p��[89��\X9�ȅ��\�9�ʅ��\�9�̅��\�9�΅��\8:�؅��]x:�څ��]�:p܅��]�:Pޅ��]8;0腵�^x;ꅹ�^�;셽�^�;��^8<��_x<��_=P ��`�=��aX>�(��b�>p/��c�?�>��c�?H��d�?�I��d(H�K��dhH�M��d�H`O��e�H@Y��e(I [��ehI]��e�I�^��e�I�h��f(J�j��fhJ�l��f�J�n��f�J�x��g�J�z��g(K�|��ghK`~��g�K@��h�K ��h(L��hhL��h�L��i�L��i(M��ihM`��i�M@��j�M ��j(N��jhN��j�N ��k�N@��k(O@��khO ��k�O��k�O�Ȇ��l(X�ʆ��lhX�̆��l�X�Ά��l�X`؆��m(Y@چ��mhY ܆��m�Yކ��m�Y�߆��n(Z�醤�nhZ�놨�n�Z�n�Z`�o([��oH[��oh[��o�[@��o�[��o�[��p\P ��pH\��px\� ��p�\��q�\��q�\`��r]�)��r(]p-��rh]�9��s�]>��t�]M��u�]�^��v�]�o��w^��x(^p��zH^ଇ��{h^��|�^�؇��}�^0��~�^@��^��̀_��(_��H_@��h_@-��_p<��_�K��_`X��_�[��օh�h��ֆ(h`x��ɇHh��hh��h ��҉�hP��Ŋ�h��hp��i��(i0��Hi0��hi`��i��iP��Џ�i��i��jP��(j�*��Hj�9��hj=��j@L��j�X��j`\��j�_��k k��(k {��HkP��՘hk��ș�k@��k��kp��k@��lp��(l��Hl��hl0��l��ٟ�lP��l� ��Ơ�l��m��(m@-��Hmp<��hm0I��m�L��m`Y��m0i��֦�m`x��ɧn��(n��ߨHn ��ҩhn��n��Ϊ�n��np��ޫ�n��o��ޭ(oP��ٮHo��ho@��ү�o��ڰ�o��ձ�o�(��в�o@8��x<��(x�K��Hx�X��hx�\��x�h��x@l��x�|��xp��y ��(y`��Hy��hy`��y��y`��y��y��z��(z� ��Hz��hz��z��z`.��z=��z�K��{�X��({�\��H{�i��h{y��{��{p��{p��{ ��|��(|��H| ��h| ��|��|P��|��|��}��(} ��H}P��h}+��}�9��}�=��}`L��}@Y��~P]��(~`i��H~`m��h~�|��~��~@��~0��~0��P��(��H@��h��(��H� (��h��.��=��pJ��ȈpN��舐[��j��(��y��H�0��h�0��П��ȉ��Ю��(�p��H��x��@��(��H��ȋ`�� H��]��<threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_86_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64llvm.nvvm.associate.access.property.p0i8.p0i8llvm.nvvm.memcpy.async.p3i8.p1i8__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS_Z17__reduce_add_syncjj__reduce_add_sync_unsigned_impl_Z17__reduce_min_syncjj__reduce_min_sync_unsigned_impl_Z17__reduce_max_syncjj__reduce_max_sync_unsigned_impl_Z17__reduce_add_syncji__reduce_add_sync_signed_impl_Z17__reduce_min_syncji__reduce_min_sync_signed_impl_Z17__reduce_max_syncji__reduce_max_sync_signed_impl_Z17__reduce_and_syncjj__reduce_and_sync_unsigned_impl_Z16__reduce_or_syncjj__reduce_or_sync_unsigned_impl_Z17__reduce_xor_syncjj__reduce_xor_sync_unsigned_impl7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!|z!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQl� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@�� :�� ށ��a~��!�a��@�l X6�,�$� �b��X�@0 �f� �3b��x�@@ �� e ��H5��=�C;�A8��;�=�=��;��<�=�C8��<�;��;�C9��5�� !�!��A��!܁ʁ� c@��H�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9� �`a��HI�`� �A&�1ad�$�@�0�2af$K3�ƙP��<�@ah�DɄ��H�iB@M�h�T�`� �"�gB`M*�� 2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��2TP�� 5"PA1� 1��8F�"J�|�� Ȍk�)��ɧ��jPP4(�4��&_8��F�a ( 6��&� ��@ep� �c�)�� @" (��F�"J�|J8��FPP:(��i�(a��pDP@�� (��i�(a�)�"P@�Ѓ (��#��&_:�h�Q�G��8G�"J�|J:�h�Q�G��F�"J�|�p��FB}��>��BΑ��&��'�l$��G�B�i�(a�É� }��G!飏>h�s�)��ɧ�É� }��G!飏>ȑk�)��ɗ'�pD��G}��>��i�(a�)�p� GD}��G)&飏>� Jn��&i��c�)��ɧ�f�Q@2�4�E�"J�|��l$@8��i�(a�)��l$@>�i�(a�f��B�F)D�r�4E�0�H3�(�)��1�Q��SH3�(�*��1�Q��/Nc�(0��@�\#M%L>�8�Q��"�b�r�4E�0��60��`�r�4E�0��4��(�-@�#M%L��4ƍ0��`��5�Q��S��70��B�b�P/�HSD ��8��#�0��`�0�HSD �O-Nc��0�(�0`�4��5�Q��L5 �)�1��1�HSD �_,0�H(��2$SL1��5�Q��L7�)��B0SL1�P3�HSD �_,0�(��b�)�3�SL1�A�Q��/Nc,Nc�(0��@�\$M%L>�8��8�Q��"TcYs�4E�0��60��`es�4E�0��4��4��(�6@�$M%L��4��4ƍ0��`��E�Q��S��70��B�c��8ISD ��8��8��#�0��`�P9ISD �O-Nc,Nc��0�(�9`��9�Q��,D�u��:��9�Q��,��u�)�;�C�#M%L��P�B�hPGu �:�C�#M%L��P�B��PGu�)=ꨣ�:TO��O!� h�RT@��b�"�T��PL1TP�Q�*C1�C ��b��ʠ�:PCTL�)�!��IT�b�1�(*D1��YT�!� QL1�C�bz�8*C1C�Au�!�ʠ�:�HeP�u��2`�dRPB��<T�T0�X*� `�T ( ��$S)�QGu � ��:�C6Bu��p*�@:��u ��`�:�S�1��@B�A �ʠN1�Q�N�AJJ�P@KA� ��"P=U�+��p�TF@�BTZ�U��*FU�V�(�b��*B1�W�0��*�:�$Vh��*�Y�G�@�Ң5��Z��Q�;ܢ��"�E��.��Ex�)E8E/ڋ�)6}�_�F��#=��`DR�C��h�"L�)F��ƨ1R��<F��6E�F2��<D�A��2J��dF��3Ex�3��)�;�he��H�2\QDZ�@�2�W(�V�+�Z�T+�n�A��ኸ!��p]�le��.NH�2\��)D[��W��V�+Ӄt+��B��&!��pc�pe��1UH�2\!�6D\�x�2�W �!W�+2K��+��=C̕��!��te�bH�2�:D]!Ⱥ2T�]!�H�2T��]!�ȻB�: x��>� ��(b�*�E@�J��W�ԡ��Z��!��[$��E��>��E��H�"ħ�g(�}�E@0 �� F�>è�"ܗX��c�`��4B1��BCSL1�AB�hB#��N*��,,�@⊵�4,�n��"�>� �q]}��Gb!��H�b\��X�+�2�2�W�b1��T,��X�+룏>� qc@0��H�B\!��"�b�!�qg�SL1�c!��LuHG,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �a��0;��{� �e�p�c0��H�`��q�b0�($r�8j1�`��`�@��\0�[ ��L..��'�#��v@~�ơ��@;��` �|�q�b0�(�r�8z1�`$v�`�@�ڿ\0�_ ��b��@;��r�8�1�`d�A.�8�~5��` ��`�@��v@��r�c0��Y@�| ��x@ȡ��@;�qd9�1�`T��+�9�FPp�8�` �P, �`�@��I��#��v@�)��q�c0�hOW0�u �� @;��\�8�1�`TƂ+�;��Yp�x�` �Ph�`�@�zq��#��v@�.��q�c0�h�W0�y �� A��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��m�@�f��q�`�4�� @�d�x��!R�P:p0D�_b�H!LD�@�I��HP"�1!�C�:&$|�`�t�Ĵ4�t��!�:�Hp0D�g"* �H,MH(@�ɧ i%P"�51�C��&��`�d�E4�ԛhn��!�}�N0D2q�� H4N@`@��̉g,"�9�C$R' � `�D넴�̝x��!R�P\0D*y��H5ODw@��H/"�=!�C��'$� `��4�ܟ�N��!��j0Dz�Bj �HaXL�@��i6"�bA�C�],�� `��E��H��~��!RN�x0Dz��*�H�Y\�@��7��="�g!�C��,&� `�$�U�̴��!��}0D��B��H�Z@�@�iW��>�"�j!�C$x-�� `�į��T�`��!��}0D��Hl[0�@� o'?"�n��C$�-�� `��з��!R�S~0Db�"��H=\L�@� �'?"�q�C$W.�� `��عp��!�>P~0DR��H8]@�@� ��)?"�uQ�C$�.�� `�4�E��`��!��~0D��H<^0�@� ��?"�y��C$R/�� `�$��X��!Һ�0D��b �H _$!@�)�%D"�|��C$�/�� `��E�@��`B��!��H�0D:�� H�_$!@��?"��C��/� `�d��"@��HB��!� H�0DrA#��H7h �@��$D"��q�C�74� `��"@�@�8F��!�)ň0DbE��H�h%@�I��D�"��q�C$s4� `��R"��@J��!�>E�0D�G#)�H(i%@��% �D"٤Q�C��4� `�4��"@��@V��!�[ʊ0D�Ki�H�i$-@�I5��D"ŦA�C��4� `�䛆Q"@��HR��!R{E�0D�OC��H�i0@�)D��D"��A�C$5 `�t��!#@�z�PV��!ҙƊ0DjS)�Huj%@�IO��D"��A�C�R5� `�ԪFQ"@�d�@J��!��E�0D�U�(�H�j%@�I] �D"ɫQ�C�{5� `��FQ"@�H�PR��!��F�0D�Y�(�H;k%@��g �D"��Q�C$�5� `�D�FQ"@��H��!��0DB]cY�H�k+@�Iw �F"��C$�5� `�d�FQ"@��HJ��! F�0DRa)�H2l%@�醍�D"��a�C$36� `��ƆQ"@��HJ��!�F�0D"e#)�H�l%@�i� %E"��a�C�q6� `��FQ"@��8J��!�=D�0D�g)�H#m%@� � �D"ɴq�C��6� `�4��Q"@��hR��!�]I�0Dm�I�H�m()@� � 'E"��C$�6 `��a#@�H�HJ��!R�F�0D�qC��H=n8@��̍G"��C�;7� `��Q"@�L�@J��!��E�0D�u#��H�n8@�I� G"��q�C�y7� `��R"@�T�HR��!��E�0D�y#��Hao:@��$E"��Q�C��7� `��F�"@�B�HJ��!��F�0D�}#)�H�o%@��D"��a�C$�7� `�t��Q"@��HJ��!� F�0Db��H<x$<@��G"��q�C$=<� `��P"@�F�0J��!R*C�0Db��H�x>@��D"��1�C�_<� `�T��P"@��8|��!�;��0D��C*�H$y8%@�I&�D"��C$�<� `�ė�Q"@�F�XJ��!�kH�0D��)�H�y,%@� >��D"�灔C�=� `�Ģ�Q"@�d�h��!ҏ��0D:ӣ!�Htz�@� O�>"��C$T=3 `��Ga&@�j�@��!R��0D�1�H�z f@�i^�L"�끘 C��=3 `��b&�f�@R��!��0Dۃ �Hg{ d@��mL"��A�C��=�� `�4�r�V�@��!��0D��1�H�{f@��|�L"� C$�=3 `��b&�L�@��!��0D��I�H>| b@�ɌBL�"�� C$9>0 `��ǃ&�R�`��!R-�0D"�A �H�|<h@�韏M"�� C��>4 `��G�&��h��!�j��0D��Q �H�}<h@��M�"�� C$?4 `��ǃ&��`��!R��0DB��A �H�~<h@�i�M�"�� C$?4 `��ǃ&�r�`��!R��0D��A �H�8h@��M�"�� C��?4 `�$�ǃ&�Ba��!R"�0D�A�A �Hh�@C$<D��!)"�@��f `��&��q��!R?"�0DrI�a �H=�@n@��-�M�"�%r� C$�D 9 `�t��'8�\)��!�l"�0DZO� �H� v@�i?�N�"�(�� C�E0< `�4�H"&@��I��!R�"��0D�S$ �H��`x@��W�DL�"�+�� C�xE1 `��H"&@�F��!��"��0D"[$ �Hm�$b@��n�DL�"�-�� C��E� `�ľ�"@��yB��!� #O�0D�a��Hx�@!@� �(D�"�2�C$tF� `��'�hq��!�X#��0D�k�� H~�8~@��O"�7�� C��F? `��'@�ny��!��#�0Du� �H��8�@��GT�"�;� C$}G&�0D2��a�H5�4l@�I,�M"�e° C$�L6 `��I�&@�p3y��!�y&�0D��a�H$�8l@�IF��M"�i� C�7M6 `�ԧI�&@�T5y��!R�&�0D:��I�H��0�@�IdM"�l"� C$�M.4 `��Ʉ&@�h7��!��&�0D��B�H3�Lh@�Ɍ�M"�qB� C$XN.6 `�t�I�&@��9��!�L'��0D"��R�H{�Lj@�ɵ+U"�w"� C$�N�T `��ɔ*@��<��!��'�0Dj��B�H�\h@�IߓM"�|2� C$�O04 `�� &@�`?��!��'�0D��C�H2�Ph@��M"��R� C$YT06 `��&@�@R��!�M*��0DRKS�H��Pj@��6M"��"� C$�T.4 `�T�ʄ&@��T��!��*�0D�U�b�H�Tl@��_M"Ѭ"� C$�U.4 `�Էʄ&@�`W��!��*�0D�_�b�H1�Tl@�ɌM"�"� C$WV.4 `�T�ʄ&@��Y��!�K+�0Dk�b�Hy�Tl@�ɵM"��"� C$�V.4 `��ʄ&@��\��!��+�0DRu�b�H�Tl@��M"��"� C$�W.4 `�T�ʄ&@�P_��!��+�0D��B�H'�Lh@�I�k `��0D�["��2-@��q��H%�HC��\��!�[.�@�I4��M"��B� C��\07 `�T�K�&@�tt��!�.�0D��B�H��Lh@��]�M"��2� C��].6 `�Ķ�&@�@w��!��.�0DB��B�H��Lh@�I��M"��B� C$>^.6 `�4�K�&@��y��!R?/�0Dr��B�Hd�Hh@�ɮ��M"��"� C��^,6 `�4��&@�L|��!Ҏ/�0Db��B�H��Hh@��֗�M"��2� C$~_,6 `�$��&@��~��!�/�0Dr��B�H�Lh@� ��M"� 3� C�0d.6 `��&@�X��!�82�0D�G�B�H+�Lh@��'�M"�%C� C$�d.6 `��L�&@��QV�!�2��0D2Q�Z�H<�Lr@��M Np"�*� C$Ze&8�`�t�$'@���!�2Y��0D [&��Hr�d��iu�)"�.�� C$�e�8 `�T��*@<�z��!3]��0D�e�{�H��|�@�饙0x�Hp�x�@�ɵ��"�7��C$g�Z `�t�W�F�1�W `��C��g�� `��L�@@�ʞ�b��!�3V�0Dr}F��H��h�@�I�'V"�`s� C$4l�� `��d+@H�V��!�.6�}0DZ�f��H��DC��l&\ `�$�ͥ+@H��p��!k6�0D��H��`�@� FW"�is� C�?m.\ `�d��+�ҵq��!ҽ6��0D*��H;�`�@�in��W"�n3�C��m&p `��.�L��!7�0D��H��\�@�i��\"�s3�C��n,q `�4� &.��!�l7�0D*�f�H��X�@��\"�y��C$>o0p `�D��.�ֽ��!��7�0Dj�f!�H8�,��i��\�"�}��C��o��!�7�0Dz�b�H��@�@��H^"��C�t| `��γ/��С��!�(:��@0DrEgI(�H��,C$rtr `��N2�@҉H��!RL:��0DKGR�Hu�@-@��4f"զc�C��t�� `�d��r3�Bԉt��!��:Q�0DSG�Hu�0�@�)T�\"��ØC�\us``�T��.L��a �H��0�@�)e��]"��C��u$y `�Է�T/�^ׁ��!��:о0Dz_��H$�0�� &]�"��@��a��!;R�0D2eG"�H��L'@�i��Ie"��C��v"� `��53��aj��!�h;��0DzmG��H��H=@�i�� l"��ÐC$wr `��c.L��a�� !��;L�00Du��"��ÐC$]w"v `��N�.��ݑ��!��;R�0D{��Hp�@�@� ��_"վØC��w�t�`��0D��!�H%�H!@��d"��3�C$>|&� `��O�2��`��!R<>��0DBɇ��H4�07@� ,Ig"��#�C$�|&� `�t�C.��a��!�{>��00D�χ1&�H'�0��iF `��C.��!�>ѻ0Dj�G��H��H�@�)^_"��C��}* `�d�c.P��a� �!��>@�Iu�\"��#�C$�}$� `��t2�Z��T��!?Q�0D��'��H��\3@�)��f"��C�x~$� `�4ҏ�3�j��!Z?�0D��!�H��0��\�"��äC$�~��!|?�0D��'b�H-�D�@��I^"��#�C$Q | `�$��/��!ҽ?��@0D��I(�H/�0C��r `�t��2��H��!��?��0D��gR�H��D-@�I�f"� t�C$5�� `�d�s3�D�t��!-BR�0D2Gh�H�1�@��$!�\"�$ĘC��s``��.L��b �H� 1�@��5��]"�&�C��$y `��T/�f��!ҘBо0D�S��H� 1��V!&]�"�+�@��b��!�BR�0DRYH"�H<M'@��m�Ie"�.�C�م"� `�D��53��bj��!� C��0D�aH��HeI=@�鎡 l"�2Ĉ C$[�� `�$�C6T��b�� !�>C��P0D2i�A�H2 1b@�ɧ!Im"�5$� C$ц&� `�T��7��!R~C�0DRqȲ�H`1d��!�l�"�9Ġ C$R�� `��Є7��!��C�0DRy��H>I�@� �!It"�>��C$݇� `�d��:��T��!R F�0D��r�Hl1b@� #Fl"�bĐ C$X��`�ċc6T��1b��!=F��0D�HR�H2Il@��,��m"�e4� C$Ռ"� `��QD7��3��!��F�`0Drш1,�H:1h@��L#Fl"�i4� C�R�&� `��7��5��!�F�0D��H�Hda�@��o#�t"�nĠC$ڍ&� `�t��:��7�\��!�G��0D��Hf1d@�I�#�l�"�qĘ C$U�� `�d�#6��9��!=G�0D2�hr�H9Mp@��Hn"�u� C$׎,� `�D�C6X��;b��!G�0DB��H4Mx@��̣Io"�yD� C$W�(� `�$�:��=�D��!��G�0D��1�Hj1�@��Iu"�~4�C$ޏ(� `��#6��?b��! J�P0D�A�!*�H`(1f@�� %m"�Ĉ C$R�$� `�d��6��Q��!�>J�0DrI)��H?)Et@� .%�n"��Đ C$ٔ��`��6��Sb��!�}J�0DJQi��H<*Q|@� N%�o"��$�C$[�$� `��F:�@VbL��!�J�0D�YiR�Hk+M�@��o%�u"ɮd�C$�� `��jX��!K�}0De�� $�H�,Y�@��%v"��C$��,� `��ג;�d[�`��!zK��0D�o�� H-.Y�@�i�%�O�"ݹ��C�W�8�@`��f;�F^�t��!��K� 0D�{ ��H�/a�@��%Lw"��C��0� `��+��p��!�+N��p0D��H�8}�@�I&��>"��C$Ԝ�� `��w@��s��!R�N�}0D��H�:u�@��\��>"��C$�� `��w@�`w��!��N�}0D"��H;<}�@�� wY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B)b0E�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��b0��b0��0��0��0��b0��0�`�0��0� �0��b0�P�0��0��0��0��b0��0�P�0��0��0�@�0��0�!1��$1��%1�`(1��)1�+1�`-1� "z1��b0� "{1� "�1� "�1� "�1� "�1� "�1� "�1��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�3$6�@3%6��3(6��3)6�6*6��.26�/36�0/46�`/56��/66��/76��/86� 2=6�P2>6��2?6��2`6��2a6�3b6�@3c6��3f6��3g6�6h6��.p6�/q6�0/r6�`/s6��/t6��/u6��/v6� 2{6�P2|6��2}6��2~6��26�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3 7��3!7�6"7��.*7�/+7�0/,7�`/-7��/.7��//7��/07� 257�P267��277��287��297�3:7�@3;7��3>7��3?7�6`7��6h7��b0��6i7��b0��6j7�7k7��b0�07l7�P7m7��b0�p7n7��7s7��7t7��7u7��7v7�:w7�0:x7�P:y7��:|7��:}7��:~7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6"� ��#oC��Bܔ`0�07&� �� #w��B޴`0��7.� ��#��B��`0�0:2� ��#�c��B��`0��:8� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,T�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,D�� 6C�A8 ��\Ə`��4M��4Ma A,t�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,4�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,$�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,4�� 6C�A8 ��\Ə`��4M��4Ma A,d�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� d��\֏��4Ma A,T�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma Ĉ�� ܇� �pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` ��BpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` �qBpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a �� B�@��Ə֏��P��@��a [A,@��EzAPF�}@�^�T�xAPJ��^��}��}�` �� #�`p$�)Ȉ�� ܇ BpApqQA+\ �f5�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D&�PNS�f�`NS�&��T��`NSյ&��T�%&�շ)&��T5n� ��4o�`� �,n�`4n��`�A`� �,�}[a!�n� �,��e�!8M�߆�`NS�&@3Դ-&��T�m� ��4��[c!�a �� ~ �� B�@��Ə֏�`��#�,>Ra �� ܇~ A�@��Ə֏@��a [A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}�` ��#�`p$�)Ĉ�� ܇ BpApqQA+\ �f5�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� �,˭�`4n� ��4�m� ��4Un� ��4o� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� �,�9&��T� a �� G�B�@��Ə֏��P��@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��}��}��\��}�` � #�`p#�}�� D�p5�A85�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D6�@̀�`B!8MeaB!8M�[bB!Q}�aB!8M�`!8Me� ��4U]�b�!8MU��`��ܷY&��Txo� ��4U��>B?BX?B Xa!�n� �,n[e�!˂�v�`NS�&@3Դ-&��T�m� ��4��[c!�a �� G� �� ܇�B�@��Ə֏�`��#�,>Ra �� A�@��Ə֏@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}��\��@@��}�` �#�`p#�}�� D�p5�A85�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� ��4�m� ��4Un� ��&��T�M&�,�n� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� ��4�}[c!�a Ĉ�� pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}�` �� #�`p%�)Ȉ�� BpAp2��2B0K T�p2��2D0K@T��pqQ-\ �fA֙��0H�b� �tD��0�`� ��4�m� ��4Un� ��4o� �,�}[`!8Me� ��4U]�b�!8MU㶏ƏF�`��#�Y&�,8o� �5m� ��4�m�c!�b� ��4��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}��\��}�` � #�`p$�}��pCh0��,�0P�vH��pC@h0�0�,1P�|00�� D�p5�A8A֙��0H�b� �tD��0�`� ��4�m� �,�}�aB!8M�[aB!8M�[`!8Me� ��4U]�b�!8MU㶏Ə֏F�`��Y&�,8o� �5m� ��4�m[c�!8Me��@��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a � ��}�_ d��Ə֏�0�`��,a aA,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� ܇BpAprbB� � #�`p*�}�%� D�p5�A8<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� ��4Un� ��4o� �tD�m� ��4�!&��Tum� ��4U�g�!8M�۶�`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�W d��Ə֏�0�`��,��a aA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� ܇BpApqaB��#�`p*�}�%� D�p5�A8<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�[bB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ d��Ə֏�0�`��,a `A,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� }�E�9� \�\��}�` ��'��l�@<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� �tD�m� ��4Un� ��4�!&��Tum� ��4U��aB!8M��f�!8M��ƙ`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�X d��Ə֏�0�`��,��a `A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� }�E�9� \�\@�@��}�` ��'��l�@<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�`!8Me� ��4U][bB!Q}�b�!8MU릘`NSո1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p+�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��V� 8�&@3�6�PNS�V�PNS妘PHGT�v�PNS��@NSYb�!8MU׶�`NSո}&��Tym� ��4��>B?BX?B �a!�n� �,��h�!8M��0�`� ��4UޛeBýbB4CMc�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p(�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��F�@̀�`B!8Me[aB!8M��bB!Q}�aB!8M�[`!8Me� ��4U]�b�!8MU��`NS�q&��T8o�`�A`��`� �tD�i&�,xn� ��4U��c��&��Tyo� 0�n� �5m� �c�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� ~ �p1��Q�D8��0�c�`�A`��&��T��#�,>b� ��4a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� G� �p1��Q�D8��0�c�`�A`��&��T�&��T�`��a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}e�!8Mu�&�0�c� ��4��d�!8Mu��`��ؼ&@3�e�!8M�Ӗ�`NS�a&��T�m� ��`B!8Me[aB!8M�a4n� ��4o� ��4��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�a!�n��ÿ8v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}�d� 0�U&��T7o� ��4Սe�!8Mu��`��ؼ&@3�e�!8M�ӆ�`NS�e&��T�m� �a4n� ��4�m� ��4Un� ��4�&��T��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�?��D��/�&��v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��Ə@��a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a ��A� �� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �ȿ�� #�,a �� A�@��Ə֏@��a ��A� �Ⱦ�� #�,a ��A� �H�� #�,a ��A� �Ƚ��Ə@��a ��A� �G��Ə@��a �� A�@��Ə֏@��a �� Ј� A�@��Ə֏@��a ��A� �Ȼ��Ə@��a ��A� �H��Ə@��a ��A� �Ⱥ��Ə@��a ��A� �H��Ə@��a ��A� �ȹ��Ə@��a ��A� �H��Ə@��a ��A� �ȸ��Ə@��a �� A�@��Ə֏@��a ��A� �ȯ��Ə@��a ��A� �H��Ə@��a ��A� �Ȯ��Ə@��a ��A� �H��Ə@��a ��A� �ȭ��Ə@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏0@�c��,a ��A� �H��Ə@��a ��A� �Ȫ��Ə@��a ��A� �H��Ə@��a ��A� �ȩ��Ə@��a ��A� �H��Ə@��a ��A� �Ȩ��Ə@��a ��A� �H��Ə@��a ��A� �ȟ��Ə@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏�@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H��Ə@��a ��A� �Ȝ��Ə@��a ��A� �H��Ə@��a ��A� �ț��Ə@��a ��A� �H��Ə@��a �� H� B��@��Ə@H�Y?!<Sa �� Ț B��@��Ə@H�Y?!<Sa ��A� �ș��Ə@��a ��A� �H��Ə@��a �� A�@Ə��4֏@��a ��A� �H��Ə@��a ��A� �ȏ��Ə@��a �� A�@Ə��4֏@��a ��A� �Ȏ��Ə@��a ��A� �H��Ə@��a ��A� �ȍ��Ə@��a ��A� �H��Ə@��a ��A� �Ȍ��Ə@��a ��A� �H��Ə@��a ��A� �ȋ��Ə@��a �� ɋ A�@��Ə֏@NS�?!�a � �� ɋB�@��Ə֏@NS�?!�a �� A�@��Ə֏@��a �� A�@��Ə֏@��a "D,��}��\`��#�``&-v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� I A�@��Ə֏@��a � �� B�@��Ə֏@NS�?!�a �� A�@��Ə֏@NS�?!�a �� ܇~ A�@��Ə֏@��a �� ܇~ A�@��Ə֏@��a "D,��}ğ\`��#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� A�@��Ə֏@��a �� z A�@��Ə֏@��a � �� {B�@��Ə֏@NS�?!�a �� { A�@��Ə֏@NS�?!�a �� y A�@Ə0T�b�0!I֏@��a �� Iy A�@Ə0T�b�0!I֏@��a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a �� Ix A�@Ə0T�b�0!I֏@��a �� o A�@Ə0T�b�0!I֏@��a �� Io A�@Ə0T�b�0!I֏@��a �� n A�@Ə0T�b�0!I֏@��a �� In A�@Ə0T�b�0!I֏@��a �� m A�@Ə0T�b�0!I֏@��a �� Im A�@Ə0T�b�0!I֏@��a �� l A�@Ə0T�b�0!I֏@��a �� Il A�@Ə0T�b�0!I֏@��a �� k A�@Ə0T�b�0!I֏@��a �� Ik A�@Ə0T�b�0!I֏@��a �� j A�@Ə0T�b�0!I֏@��a �� Ij A�@Ə0T�b�0!I֏@��a �� i A�@Ə0T�b�0!I֏@��a �� Ii A�@Ə0T�b�0!I֏@��a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� _ A�@Ə0T�b�0!I֏@��a �� I_ A�@Ə0T�b�0!I֏@��a �� _ B�A8֏0T�b�0!IƏp�a��,a �ȈA� �I^� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a �ȈA� ��]� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a ;J,@ �� a��pC�`0�0�� ^%�1K�7�2F0bp v�W��#�``'|�,�,�1P�D!��@�z(�p�$��A8F�`��Խ&��`B!0OM[b�!8̀Ӧ�`N3�&�,7m��&�,wo� ��<vm� ��4��#��#��#� &�,a 9J,@ �� ЕQ�$� E��,�@#�` &u��p�Q��,Ca#�` &v�$�ǈ�� ]-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 9J,@ �� Q�$� E��,�@#�` &p��p�Q��,Ca#�` &q�$�ǈ�� I\-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3�@��a!8�`� ��<�&�,vm� ��<��Əf�`��ػ&�Թ%&��vo� ��4��a 9J,@ �� Q�$� E��,�@#�` &k��p�Q��,Ca#�` &l�$�ǈ�� [-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 5J,@� ��,C #�`"cE�$� ��,�@#�`"de� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`">E�$� ��,�@#�`"?e� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'9E�$� ��,�@#�`�':e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'4E�$� ��,�@#�`�'5e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP �"S�,�1bP �"S�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP ��R�,�1bP ��R�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a � �X�@��Ə@0�@X?!8Ma � �T�@��Ə@0�@X?!8Ma � �`�@��Ə@0�@X?!8Ma �� H A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� - A�@Ə0T�b�0!I֏@��a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a �� , A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� ,94A�@֏0T�b�0!IƏp�a��,a �� ,74A�@֏0T�b�0!IƏp�a��,a �� ,54A�@֏0T�b�0!IƏp�a��,a �� ,34A�@֏0T�b�0!IƏp�a��,a �� I B�A8֏0T�b�0!IƏp�a��,a �� B�A8֏0T�b�0!IƏp�a��,a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �ȈA� � ��Ə@NS�>�D�`��,a �ȈA� ȉ��Ə@NS�>�D�`��,a �ȈA� � �� @�@ �0�c��4��#��T�&��T��#OD֏@��a �ȈA� ȉ�� @�@ �0�c��4��#��T�&��T��#OD֏@��a ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa �b�d�@��<a �a�d�@��<a ��d�@��<a ��d�@��<a �b�p�@��pD� �a �a�p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a A, �\`d�`�n@��`tdld{@�;�;#�`�*��A8��0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#�`P��A8��pD� �a �� G�B�A8֏`8E!��# )Ə@�Da ��A� �J�� a �� Ə0�d�2a ��A� �J��@0��a �pCPa0bp ��7�� Ap�A8Ə@<1�>��a A,@ � �� ؊�B0��Ə@<1�>��a �pCPa0bp ��7C� Ap�A8Ə@<1�>��a A,@� �� ؊�B0��Ə@<1�>��a A, 4�\W ��8�C:�C9�C8��:��8́��} #�` a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұX?�!EB 4��#Y`� Q��# )a A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#֏pH�P a� �� &ұ��#�>��a A,@�&!P�܈A� � � �a� �A8 �PX�0&�0H�b�CD֏P�2��@0��X`� Qa A,@�&!P�܈A� � � �a� �A8�PX�0�PHT�#&�PH��#Ə0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұ?�P��#�L��# )�0HTa A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#6�@8��`B!��?�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0F� 8� &ұ?�P��# )&�0HT`��4� H a A,@�O6!P\Ď$��B$H0�� PX�0&�PHT�#F�@8��`B!��?�P��# )֏�,@C��a #A,@�&!P�\Î�� 2b� ��7J�0�#�`�+rd�PX�06� 8� &ұX?�!EB 4�&�,�m�CD�0HT`�H�&�,a A,@�&!P�܈A� ��# �`�+y��!�&��A� �J�� PX�0f�`��Է &ұ?�P��#�L��# )�0HT`� �,a "A,@�&!P�\Î�� 2b� ��7 �0�#�`�+rd�PX�06� 8� &ұ�a�!�R�Ə0T@d��@,`�H�&a!�a #A,@�O6!P\Ď��M2b� ��7K$�hB��M�A8�PX�0F� 8� &ұb�!�R�Ə0T@d� �D��@0��X?� a�b� �,a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H��#R$�@C�b�!Q��Ԇ�@8��aBHLDH�2?�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�&ұ��a�HLDH�2��-& �H�#��`8��a FA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�PHT�#��`H��#�&(��D�D-Sm� �t,>b�H�)&�Hm� 1!Q�d�CD֏P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H�-&�Hm� 1!Q�d� ��#Y?B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a JA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�PHT�#�P8��֘`H��#�&(��D�D-Sm� �t,>b�H��# �& )Ƙ`HT�#�%&N�#V��LƏ0T@a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a MA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#��#R$�@C�b"Q��Hm� �t,>�#�&$��D�D-��#�a�!��a GA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0V��L�&��>��b� �D>�#�-& ұ��F��LƏ0T@d��@,`� �t,>�#a LA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#�-& ��Ƙ�H��#>baBHLDH�2?�P��#�L�&ұ��a MA,@�OPQsu��`n v@`@p�`�dl@r�{`�9��<��<�{sy��`n v@`�dl@q�ld {p�<�A��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0f��L�&��>��b��4� H�%&N�#>b� ��Ƙ�HT�#>R[c"��m� 1!Q�d�CD��pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�!&N�#>b� �t,>�#F��L>b� �D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0��HT�#��#R$�@C�aB"��aBHLDH�2��b�D>�#>b�CD��@0��aB"Q��XcB�D>r��c�"��>�#V��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0��HT�#��&$ұ��&4��D�D-��#Y?B1��>��b� �D>�#>b� �t,>b��Xa�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0��HT�#��&$ұ��&4��D�D-��!& N�#>�#Ə0T@d��@,`�H�&$��5&$N�#��1&(ұ��#>b� 1!Q�T��a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a PA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0Ƙ�HT�#��!&$ұ��&4��D�D-��%& N�#>�#Ə0T@d�H��# �& )v��HT�#>�#昐8��>�#֘�H��#��&8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� �_�O�H#�d� ��4UMaB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &��T&��L��#�Y`�HLDH�2�F�`NSո&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &(��D�D-o� 1!Q˴a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6�@NSY`�HLDH�2�&��L�&(��D�D-�n� ��4U�[a�!8MU��L��0<Sd� �_�O�H#�d� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a �� /A�@ƏPT�B��@0��a �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@ƏPT�B��@0��a �� /A�@ƏPT�B��@0��a ��A� ��-#�`p�0��@0��X?!�b�E,a ��A� �I-#�`;�0��@0��X?!�b�E,a �� -�hB�&�İQ�C8֏@<1�>��b�E,a �� -�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a ��A� �.#�`�;�@� 0� ��@�@֏@<1�>��b�E,a��,a ��A� ��-#�`�;�@� 0� ��@�@ƏPT�B֏@<1�>��b��,a A, $�\U�C8��;�;�C9��A #�`+��@�� 4a �� O*B�A8֏0T�b�0!IƏp�a��,a �� ܇)A�@��@0��X?!�b�E,a �� )A�@��@0��X?!�b�E,a �� ܇(A�@��@0��X?!�b�E,a �� (A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��?BQEX?!�a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,q y2"�^�4�uJ�`�4�L�b�48�O�d�5Xf[�f�5xv^�h�5��h�j�6��k�l�6�fo�n�7�vz�p�7�|�r�78��t�8Xg��v�8xw��x�8�ǘ�z�9��|�9�g��~�:�w��:`��:8p��;X��;x��;�`Ƀ��<�p̃��<�`σ��=�ڃ��>8�郔�>X��?x��@�0��@��A��B��.��C�?��D8�O��EX_��FxPn��G�p}��H��I�@��J�`��KP��K8ʄ��MX�ބ��N��턺�N��N��OH��O��O��O��O8��Ph� ��P��P��P(@��Ph��Q�`��Q��Q(��Qh��Q��R�)��R((�*��Rh(,��R�(�-��R�(0/��R()�8��Sh)0:��S�)�;��S�)@=��S(*�>��Sh*@H��T�*�I��T�*�K��T(+ M��Th+�N��T�+ X��U�+�Y��U(,P[��Uh,�\��U�,p^��U�,�_��V(-�i��Vh-0k��V�-�l��V�-@n��V(.�o��Wh.Py��W�.�z��W�.P|��W(/0~��WH/��X�/P��X�/И��Y�/p��Y�/��Y8��YH8��Zh80��Z�8��Z�80��[�8л��[9p��[89��\X9�ȅ��\�9�ʅ��\�9�̅��\�9�΅��\8:�؅��]x:�څ��]�:p܅��]�:Pޅ��]8;0腵�^x;ꅹ�^�;셽�^�;��^8<��_x<��_=P ��`�=��aX>�(��b�>p/��c�?�>��c�?H��d�?�I��d(H�K��dhH�M��d�H`O��e�H@Y��e(I [��ehI]��e�I�^��e�I�h��f(J�j��fhJ�l��f�J�n��f�J�x��g�J�z��g(K�|��ghK`~��g�K@��h�K ��h(L��hhL��h�L��i�L��i(M��ihM`��i�M@��j�M ��j(N��jhN��j�N ��k�N@��k(O@��khO ��k�O��k�O�Ȇ��l(X�ʆ��lhX�̆��l�X�Ά��l�X`؆��m(Y@چ��mhY ܆��m�Yކ��m�Y�߆��n(Z�醤�nhZ�놨�n�Z�n�Z`�o([��oH[��oh[��o�[@��o�[��o�[��p\P ��pH\��px\� ��p�\��q�\��q�\`��r]�)��r(]p-��rh]�9��s�]>��t�]M��u�]�^��v�]�o��w^��x(^p��zH^ଇ��{h^��|�^�؇��}�^0��~�^@��^��̀_��(_��H_@��h_@-��_p<��_�K��_`X��_�[��օh�h��ֆ(h`x��ɇHh��hh��h ��҉�hP��Ŋ�h��hp��i��(i0��Hi0��hi`��i��iP��Џ�i��i��jP��(j�*��Hj�9��hj=��j@L��j�X��j`\��j�_��k k��(k {��HkP��՘hk��ș�k@��k��kp��k@��lp��(l��Hl��hl0��l��ٟ�lP��l� ��Ơ�l��m��(m@-��Hmp<��hm0I��m�L��m`Y��m0i��֦�m`x��ɧn��(n��ߨHn ��ҩhn��n��Ϊ�n��np��ޫ�n��o��ޭ(oP��ٮHo��ho@��ү�o��ڰ�o��ձ�o�(��в�o@8��x<��(x�K��Hx�X��hx�\��x�h��x@l��x�|��xp��y ��(y`��Hy��hy`��y��y`��y��y��z��(z� ��Hz��hz��z��z`.��z=��z�K��{�X��({�\��H{�i��h{y��{��{p��{p��{ ��|��(|��H| ��h| ��|��|P��|��|��}��(} ��H}P��h}+��}�9��}�=��}`L��}@Y��~P]��(~`i��H~`m��h~�|��~��~@��~0��~0��P��(��H@��h��(��H� (��h��.��=��pJ��ȈpN��舐[��j��(��y��H�0��h�0��П��ȉ��Ю��(�p��H��x��@��(��H��ȋ`�� H��]��<threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_87_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64llvm.nvvm.associate.access.property.p0i8.p0i8llvm.nvvm.memcpy.async.p3i8.p1i8__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS_Z17__reduce_add_syncjj__reduce_add_sync_unsigned_impl_Z17__reduce_min_syncjj__reduce_min_sync_unsigned_impl_Z17__reduce_max_syncjj__reduce_max_sync_unsigned_impl_Z17__reduce_add_syncji__reduce_add_sync_signed_impl_Z17__reduce_min_syncji__reduce_min_sync_signed_impl_Z17__reduce_max_syncji__reduce_max_sync_signed_impl_Z17__reduce_and_syncjj__reduce_and_sync_unsigned_impl_Z16__reduce_or_syncjj__reduce_or_sync_unsigned_impl_Z17__reduce_xor_syncjj__reduce_xor_sync_unsigned_impl7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!|z!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQl� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@�� :�� ށ��a~��!�a��@�l X6�,�$� �b��X�@0 �f� �3b��x�@@ �� e ��H5��=�C;�A8��;�=�=��;��<�=�C8��<�;��;�C9��5�� !�!��A��!܁ʁ� c@��H�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9� �`a��HI�`� �A&�1ad�$�@�0�2af$K3�ƙP��<�@ah�DɄ��H�iB@M�h�T�`� �"�gB`M*�� 2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��2TP�� 5"PA1� 1��8F�"J�|�� Ȍk�)��ɧ��jPP4(�4��&_8��F�a ( 6��&� ��@ep� �c�)�� @" (��F�"J�|J8��FPP:(��i�(a��pDP@�� (��i�(a�)�"P@�Ѓ (��#��&_:�h�Q�G��8G�"J�|J:�h�Q�G��F�"J�|�p��FB}��>��BΑ��&��'�l$��G�B�i�(a�É� }��G!飏>h�s�)��ɧ�É� }��G!飏>ȑk�)��ɗ'�pD��G}��>��i�(a�)�p� GD}��G)&飏>� Jn��&i��c�)��ɧ�f�Q@2�4�E�"J�|��l$@8��i�(a�)��l$@>�i�(a�f��B�F)D�r�4E�0�H3�(�)��1�Q��SH3�(�*��1�Q��/Nc�(0��@�\#M%L>�8�Q��"�b�r�4E�0��60��`�r�4E�0��4��(�-@�#M%L��4ƍ0��`��5�Q��S��70��B�b�P/�HSD ��8��#�0��`�0�HSD �O-Nc��0�(�0`�4��5�Q��L5 �)�1��1�HSD �_,0�H(��2$SL1��5�Q��L7�)��B0SL1�P3�HSD �_,0�(��b�)�3�SL1�A�Q��/Nc,Nc�(0��@�\$M%L>�8��8�Q��"TcYs�4E�0��60��`es�4E�0��4��4��(�6@�$M%L��4��4ƍ0��`��E�Q��S��70��B�c��8ISD ��8��8��#�0��`�P9ISD �O-Nc,Nc��0�(�9`��9�Q��,D�u��:��9�Q��,��u�)�;�C�#M%L��P�B�hPGu �:�C�#M%L��P�B��PGu�)=ꨣ�:TO��O!� h�RT@��b�"�T��PL1TP�Q�*C1�C ��b��ʠ�:PCTL�)�!��IT�b�1�(*D1��YT�!� QL1�C�bz�8*C1C�Au�!�ʠ�:�HeP�u��2`�dRPB��<T�T0�X*� `�T ( ��$S)�QGu � ��:�C6Bu��p*�@:��u ��`�:�S�1��@B�A �ʠN1�Q�N�AJJ�P@KA� ��"P=U�+��p�TF@�BTZ�U��*FU�V�(�b��*B1�W�0��*�:�$Vh��*�Y�G�@�Ң5��Z��Q�;ܢ��"�E��.��Ex�)E8E/ڋ�)6}�_�F��#=��`DR�C��h�"L�)F��ƨ1R��<F��6E�F2��<D�A��2J��dF��3Ex�3��)�;�he��H�2\QDZ�@�2�W(�V�+�Z�T+�n�A��ኸ!��p]�le��.NH�2\��)D[��W��V�+Ӄt+��B��&!��pc�pe��1UH�2\!�6D\�x�2�W �!W�+2K��+��=C̕��!��te�bH�2�:D]!Ⱥ2T�]!�H�2T��]!�ȻB�: x��>� ��(b�*�E@�J��W�ԡ��Z��!��[$��E��>��E��H�"ħ�g(�}�E@0 �� F�>è�"ܗX��c�`��4B1��BCSL1�AB�hB#��N*��,,�@⊵�4,�n��"�>� �q]}��Gb!��H�b\��X�+�2�2�W�b1��T,��X�+룏>� qc@0��H�B\!��"�b�!�qg�SL1�c!��LuHG,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8�1�`� �a��0;��{� �e�p�c0��H�`��q�b0�($r�8j1�`��`�@��\0�[ ��L..��'�#��v@~�ơ��@;��` �|�q�b0�(�r�8z1�`$v�`�@�ڿ\0�_ ��b��@;��r�8�1�`d�A.�8�~5��` ��`�@��v@��r�c0��Y@�| ��x@ȡ��@;�qd9�1�`T��+�9�FPp�8�` �P, �`�@��I��#��v@�)��q�c0�hOW0�u �� @;��\�8�1�`TƂ+�;��Yp�x�` �Ph�`�@�zq��#��v@�.��q�c0�h�W0�y �� A��@;�S\�8�1�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��m�@�f��q�`�4�� @�d�x��!R�P:p0D�_b�H!LD�@�I��HP"�1!�C�:&$|�`�t�Ĵ4�t��!�:�Hp0D�g"* �H,MH(@�ɧ i%P"�51�C��&��`�d�E4�ԛhn��!�}�N0D2q�� H4N@`@��̉g,"�9�C$R' � `�D넴�̝x��!R�P\0D*y��H5ODw@��H/"�=!�C��'$� `��4�ܟ�N��!��j0Dz�Bj �HaXL�@��i6"�bA�C�],�� `��E��H��~��!RN�x0Dz��*�H�Y\�@��7��="�g!�C��,&� `�$�U�̴��!��}0D��B��H�Z@�@�iW��>�"�j!�C$x-�� `�į��T�`��!��}0D��Hl[0�@� o'?"�n��C$�-�� `��з��!R�S~0Db�"��H=\L�@� �'?"�q�C$W.�� `��عp��!�>P~0DR��H8]@�@� ��)?"�uQ�C$�.�� `�4�E��`��!��~0D��H<^0�@� ��?"�y��C$R/�� `�$��X��!Һ�0D��b �H _$!@�)�%D"�|��C$�/�� `��E�@��`B��!��H�0D:�� H�_$!@��?"��C��/� `�d��"@��HB��!� H�0DrA#��H7h �@��$D"��q�C�74� `��"@�@�8F��!�)ň0DbE��H�h%@�I��D�"��q�C$s4� `��R"��@J��!�>E�0D�G#)�H(i%@��% �D"٤Q�C��4� `�4��"@��@V��!�[ʊ0D�Ki�H�i$-@�I5��D"ŦA�C��4� `�䛆Q"@��HR��!R{E�0D�OC��H�i0@�)D��D"��A�C$5 `�t��!#@�z�PV��!ҙƊ0DjS)�Huj%@�IO��D"��A�C�R5� `�ԪFQ"@�d�@J��!��E�0D�U�(�H�j%@�I] �D"ɫQ�C�{5� `��FQ"@�H�PR��!��F�0D�Y�(�H;k%@��g �D"��Q�C$�5� `�D�FQ"@��H��!��0DB]cY�H�k+@�Iw �F"��C$�5� `�d�FQ"@��HJ��! F�0DRa)�H2l%@�醍�D"��a�C$36� `��ƆQ"@��HJ��!�F�0D"e#)�H�l%@�i� %E"��a�C�q6� `��FQ"@��8J��!�=D�0D�g)�H#m%@� � �D"ɴq�C��6� `�4��Q"@��hR��!�]I�0Dm�I�H�m()@� � 'E"��C$�6 `��a#@�H�HJ��!R�F�0D�qC��H=n8@��̍G"��C�;7� `��Q"@�L�@J��!��E�0D�u#��H�n8@�I� G"��q�C�y7� `��R"@�T�HR��!��E�0D�y#��Hao:@��$E"��Q�C��7� `��F�"@�B�HJ��!��F�0D�}#)�H�o%@��D"��a�C$�7� `�t��Q"@��HJ��!� F�0Db��H<x$<@��G"��q�C$=<� `��P"@�F�0J��!R*C�0Db��H�x>@��D"��1�C�_<� `�T��P"@��8|��!�;��0D��C*�H$y8%@�I&�D"��C$�<� `�ė�Q"@�F�XJ��!�kH�0D��)�H�y,%@� >��D"�灔C�=� `�Ģ�Q"@�d�h��!ҏ��0D:ӣ!�Htz�@� O�>"��C$T=3 `��Ga&@�j�@��!R��0D�1�H�z f@�i^�L"�끘 C��=3 `��b&�f�@R��!��0Dۃ �Hg{ d@��mL"��A�C��=�� `�4�r�V�@��!��0D��1�H�{f@��|�L"� C$�=3 `��b&�L�@��!��0D��I�H>| b@�ɌBL�"�� C$9>0 `��ǃ&�R�`��!R-�0D"�A �H�|<h@�韏M"�� C��>4 `��G�&��h��!�j��0D��Q �H�}<h@��M�"�� C$?4 `��ǃ&��`��!R��0DB��A �H�~<h@�i�M�"�� C$?4 `��ǃ&�r�`��!R��0D��A �H�8h@��M�"�� C��?4 `�$�ǃ&�Ba��!R"�0D�A�A �Hh�@C$<D��!)"�@��f `��&��q��!R?"�0DrI�a �H=�@n@��-�M�"�%r� C$�D 9 `�t��'8�\)��!�l"�0DZO� �H� v@�i?�N�"�(�� C�E0< `�4�H"&@��I��!R�"��0D�S$ �H��`x@��W�DL�"�+�� C�xE1 `��H"&@�F��!��"��0D"[$ �Hm�$b@��n�DL�"�-�� C��E� `�ľ�"@��yB��!� #O�0D�a��Hx�@!@� �(D�"�2�C$tF� `��'�hq��!�X#��0D�k�� H~�8~@��O"�7�� C��F? `��'@�ny��!��#�0Du� �H��8�@��GT�"�;� C$}G&�0D2��a�H5�4l@�I,�M"�e° C$�L6 `��I�&@�p3y��!�y&�0D��a�H$�8l@�IF��M"�i� C�7M6 `�ԧI�&@�T5y��!R�&�0D:��I�H��0�@�IdM"�l"� C$�M.4 `��Ʉ&@�h7��!��&�0D��B�H3�Lh@�Ɍ�M"�qB� C$XN.6 `�t�I�&@��9��!�L'��0D"��R�H{�Lj@�ɵ+U"�w"� C$�N�T `��ɔ*@��<��!��'�0Dj��B�H�\h@�IߓM"�|2� C$�O04 `�� &@�`?��!��'�0D��C�H2�Ph@��M"��R� C$YT06 `��&@�@R��!�M*��0DRKS�H��Pj@��6M"��"� C$�T.4 `�T�ʄ&@��T��!��*�0D�U�b�H�Tl@��_M"Ѭ"� C$�U.4 `�Էʄ&@�`W��!��*�0D�_�b�H1�Tl@�ɌM"�"� C$WV.4 `�T�ʄ&@��Y��!�K+�0Dk�b�Hy�Tl@�ɵM"��"� C$�V.4 `��ʄ&@��\��!��+�0DRu�b�H�Tl@��M"��"� C$�W.4 `�T�ʄ&@�P_��!��+�0D��B�H'�Lh@�I�k `��0D�["��2-@��q��H%�HC��\��!�[.�@�I4��M"��B� C��\07 `�T�K�&@�tt��!�.�0D��B�H��Lh@��]�M"��2� C��].6 `�Ķ�&@�@w��!��.�0DB��B�H��Lh@�I��M"��B� C$>^.6 `�4�K�&@��y��!R?/�0Dr��B�Hd�Hh@�ɮ��M"��"� C��^,6 `�4��&@�L|��!Ҏ/�0Db��B�H��Hh@��֗�M"��2� C$~_,6 `�$��&@��~��!�/�0Dr��B�H�Lh@� ��M"� 3� C�0d.6 `��&@�X��!�82�0D�G�B�H+�Lh@��'�M"�%C� C$�d.6 `��L�&@��QV�!�2��0D2Q�Z�H<�Lr@��M Np"�*� C$Ze&8�`�t�$'@���!�2Y��0D [&��Hr�d��iu�)"�.�� C$�e�8 `�T��*@<�z��!3]��0D�e�{�H��|�@�饙0x�Hp�x�@�ɵ��"�7��C$g�Z `�t�W�F�1�W `��C��g�� `��L�@@�ʞ�b��!�3V�0Dr}F��H��h�@�I�'V"�`s� C$4l�� `��d+@H�V��!�.6�}0DZ�f��H��DC��l&\ `�$�ͥ+@H��p��!k6�0D��H��`�@� FW"�is� C�?m.\ `�d��+�ҵq��!ҽ6��0D*��H;�`�@�in��W"�n3�C��m&p `��.�L��!7�0D��H��\�@�i��\"�s3�C��n,q `�4� &.��!�l7�0D*�f�H��X�@��\"�y��C$>o0p `�D��.�ֽ��!��7�0Dj�f!�H8�,��i��\�"�}��C��o��!�7�0Dz�b�H��@�@��H^"��C�t| `��γ/��С��!�(:��@0DrEgI(�H��,C$rtr `��N2�@҉H��!RL:��0DKGR�Hu�@-@��4f"զc�C��t�� `�d��r3�Bԉt��!��:Q�0DSG�Hu�0�@�)T�\"��ØC�\us``�T��.L��a �H��0�@�)e��]"��C��u$y `�Է�T/�^ׁ��!��:о0Dz_��H$�0�� &]�"��@��a��!;R�0D2eG"�H��L'@�i��Ie"��C��v"� `��53��aj��!�h;��0DzmG��H��H=@�i�� l"��ÐC$wr `��c.L��a�� !��;L�00Du��"��ÐC$]w"v `��N�.��ݑ��!��;R�0D{��Hp�@�@� ��_"վØC��w�t�`��0D��!�H%�H!@��d"��3�C$>|&� `��O�2��`��!R<>��0DBɇ��H4�07@� ,Ig"��#�C$�|&� `�t�C.��a��!�{>��00D�χ1&�H'�0��iF `��C.��!�>ѻ0Dj�G��H��H�@�)^_"��C��}* `�d�c.P��a� �!��>@�Iu�\"��#�C$�}$� `��t2�Z��T��!?Q�0D��'��H��\3@�)��f"��C�x~$� `�4ҏ�3�j��!Z?�0D��!�H��0��\�"��äC$�~��!|?�0D��'b�H-�D�@��I^"��#�C$Q | `�$��/��!ҽ?��@0D��I(�H/�0C��r `�t��2��H��!��?��0D��gR�H��D-@�I�f"� t�C$5�� `�d�s3�D�t��!-BR�0D2Gh�H�1�@��$!�\"�$ĘC��s``��.L��b �H� 1�@��5��]"�&�C��$y `��T/�f��!ҘBо0D�S��H� 1��V!&]�"�+�@��b��!�BR�0DRYH"�H<M'@��m�Ie"�.�C�م"� `�D��53��bj��!� C��0D�aH��HeI=@�鎡 l"�2Ĉ C$[�� `�$�C6T��b�� !�>C��P0D2i�A�H2 1b@�ɧ!Im"�5$� C$ц&� `�T��7��!R~C�0DRqȲ�H`1d��!�l�"�9Ġ C$R�� `��Є7��!��C�0DRy��H>I�@� �!It"�>��C$݇� `�d��:��T��!R F�0D��r�Hl1b@� #Fl"�bĐ C$X��`�ċc6T��1b��!=F��0D�HR�H2Il@��,��m"�e4� C$Ռ"� `��QD7��3��!��F�`0Drш1,�H:1h@��L#Fl"�i4� C�R�&� `��7��5��!�F�0D��H�Hda�@��o#�t"�nĠC$ڍ&� `�t��:��7�\��!�G��0D��Hf1d@�I�#�l�"�qĘ C$U�� `�d�#6��9��!=G�0D2�hr�H9Mp@��Hn"�u� C$׎,� `�D�C6X��;b��!G�0DB��H4Mx@��̣Io"�yD� C$W�(� `�$�:��=�D��!��G�0D��1�Hj1�@��Iu"�~4�C$ޏ(� `��#6��?b��! J�P0D�A�!*�H`(1f@�� %m"�Ĉ C$R�$� `�d��6��Q��!�>J�0DrI)��H?)Et@� .%�n"��Đ C$ٔ��`��6��Sb��!�}J�0DJQi��H<*Q|@� N%�o"��$�C$[�$� `��F:�@VbL��!�J�0D�YiR�Hk+M�@��o%�u"ɮd�C$�� `��jX��!K�}0De�� $�H�,Y�@��%v"��C$��,� `��ג;�d[�`��!zK��0D�o�� H-.Y�@�i�%�O�"ݹ��C�W�8�@`��f;�F^�t��!��K� 0D�{ ��H�/a�@��%Lw"��C��0� `��+��p��!�+N��p0D��H�8}�@�I&��>"��C$Ԝ�� `��w@��s��!R�N�}0D��H�:u�@��\��>"��C$�� `��w@�`w��!��N�}0D"��H;<}�@�� wY $2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B)b0E�8�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y bH C�9$��@��h"P2�!GȐQ�alignF&0��b0��b0��0��0��0��b0��0�`�0��0� �0��b0�P�0��0��0��0��b0��0�P�0��0��0�@�0��0�!1��$1��%1�`(1��)1�+1�`-1� "z1��b0� "{1� "�1� "�1� "�1� "�1� "�1� "�1��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�3$6�@3%6��3(6��3)6�6*6��.26�/36�0/46�`/56��/66��/76��/86� 2=6�P2>6��2?6��2`6��2a6�3b6�@3c6��3f6��3g6�6h6��.p6�/q6�0/r6�`/s6��/t6��/u6��/v6� 2{6�P2|6��2}6��2~6��26�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3 7��3!7�6"7��.*7�/+7�0/,7�`/-7��/.7��//7��/07� 257�P267��277��287��297�3:7�@3;7��3>7��3?7�6`7��6h7��b0��6i7��b0��6j7�7k7��b0�07l7�P7m7��b0�p7n7��7s7��7t7��7u7��7v7�:w7�0:x7�P:y7��:|7��:}7��:~7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6"� ��#oC��Bܔ`0�07&� �� #w��B޴`0��7.� ��#��B��`0�0:2� ��#�c��B��`0��:8� �3E �0�1À�C3Jp�0,3��03O��0@A4� ��D3T�0TA4�`��03X�0dA4à��D3\�0tA4��C`0� �0�A3d��C`0�`�0�A3h3i3j3k03l03mD3nD3o3p3q3r03s03tD3uD3v03wD3xD3y3z3{3|03}03~D3D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3�03�03�D3�D3�3�3�3�03�03�D3�D3�03�D3�D3�3�3�3 03!03"D3#D3$3%3&3'03(03)D3*D3+03,D3-D3.��L11� AL�0�D@35��M71À�M�0�D�3:��N11��M�0�D@3>��O71�M�0�Ep3b��X/1�@AL�0�E3f��Y41À�M�0�Ep3j��Z11��AL�0�E@3n��[71��M�0�E@3r��\71�@�K�0�E3v��]41ÀM�0�Ep3z��^/1��AL�0�E3~��_41��M�0�Fp3��h71�@�M�0�F�3��i11ÀM�0�F@3��j71��K�0�F3��k41�M�0�Fp3��l41�@�M�0�Fp�3��'b b b b b b b b b b b b�n�n�n�n�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n�2��t��\��F j#!7;�6�07�7�0��7��QT�6n7rC7v�7z�7~<�C<ƃ<��<�=�C=փ=��=�>�C>�>��>�?�C?��?��?�DBDDF�DJ�DNERDEV�EZ�E^FbDFf�Fj�FnGrDGv�Gz�G~L�DLƄL��L�M�DMքM��M�N�DN�N��N�O�DO��O��O�TBETF�TJ�TNUREUV�UZ�U^VbEVf�Vj�VnWrEWv�Wz�W~\�E\ƅ\��\�]�E]օ]��]�^�E^�^��^�_�E_��_��_�dBFdF�dJ�dNeRFeV�eZ�e^fbFff�fj�fngrFgv�gz�g~� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,T�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,D�� 6C�A8 ��\Ə`��4M��4Ma A,t�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,4�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,$�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� d��\֏��4Ma A,4�� 6C�A8 ��\Ə`��4M��4Ma A,d�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,d�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,$�� d��\֏��4Ma A,T�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,t�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma A,�� d��\֏��4Ma A,�� 6C�A8 ��\Ə`��4M��4Ma A,�� 6D0�B�@ ��\Ə`֏d��4M��4M��4Ma A,D�� A�6E0��lCP��4M&��4M�X��\Ə`֏d6��4M��4Ma Ĉ�� ܇� �pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` ��BpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a Ĉ�� pC`f0��,�0K0TD�B�@�0�c�`�A`��4��#�,a YE,@��P#EzAPF�}@�^�T�xApr2b` �qBpAp r2b` ��BpAp2��2B0K T�p2��2D0K@T��pqQ�*\ �f1�`NSٽQ&$@#��]&ұX`� �E&$@#��=&��Twm� ��4Սd� 0�5&�,6o��f�`NS�U&��T8m� ��`4n� ��4�m� ��4Un� ��4o� ��4�&��T5n� ��4U��>B?BXa!�n� ��<wn� ��4�M�b�!8Meۖ�@��c�!8Me�a �� B�@��Ə֏��P��@��a [A,@��EzAPF�}@�^�T�xAPJ��^��}��}�` �� #�`p$�)Ȉ�� ܇ BpApqQA+\ �f5�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D&�PNS�f�`NS�&��T��`NSյ&��T�%&�շ)&��T5n� ��4o�`� �,n�`4n��`�A`� �,�}[a!�n� �,��e�!8M�߆�`NS�&@3Դ-&��T�m� ��4��[c!�a �� ~ �� B�@��Ə֏�`��#�,>Ra �� ܇~ A�@��Ə֏@��a [A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}�` ��#�`p$�)Ĉ�� ܇ BpApqQA+\ �f5�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� �,˭�`4n� ��4�m� ��4Un� ��4o� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� �,�9&��T� a �� G�B�@��Ə֏��P��@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��}��}��\��}�` � #�`p#�}�� D�p5�A85�`NSٽe&ұX`� �A&��T7n� ��4խ�d�!8Mu��0�`� ��4��f��D6�@̀�`B!8MeaB!8M�[bB!Q}�aB!8M�`!8Me� ��4U]�b�!8MU��`��ܷY&��Txo� ��4U��>B?BX?B Xa!�n� �,n[e�!˂�v�`NS�&@3Դ-&��T�m� ��4��[c!�a �� G� �� ܇�B�@��Ə֏�`��#�,>Ra �� A�@��Ə֏@��a ZA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��@��}��\��@@��}�` �#�`p#�}�� D�p5�A85�`NSټe&ұX`� �E&��T7m� ��4�me�!8Mu��`��ػ�#@3�e�!8M��v�`NS�a&��T�n� ��4�m� ��4Un� ��&��T�M&�,�n� �YbB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o�`�A`��`� �tD�U&�,�m� �5m� ��4�}[c!�a Ĉ�� pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}�` �� #�`p%�)Ȉ�� BpAp2��2B0K T�p2��2D0K@T��pqQ-\ �fA֙��0H�b� �tD��0�`� ��4�m� ��4Un� ��4o� �,�}[`!8Me� ��4U]�b�!8MU㶏ƏF�`��#�Y&�,8o� �5m� ��4�m�c!�b� ��4��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a Ĉ�� G��pC�f0��,�0K0TD B�@ �@N3`� 0��#��#��#��#�,��#�,a sE,@��P#UzAP��}@!�^��xAP��^��}��}��\��}�` � #�`p$�}��pCh0��,�0P�vH��pC@h0�0�,1P�|00�� D�p5�A8A֙��0H�b� �tD��0�`� ��4�m� �,�}�aB!8M�[aB!8M�[`!8Me� ��4U]�b�!8MU㶏Ə֏F�`��Y&�,8o� �5m� ��4�m[c�!8Me��@��c�!8Me��`NSݸI&��T�n� ��<6�`NSݼq&��U&��T8n� ��e�!8M��F�@̀�bB!Q}f�!8M�׶�`NS�&��`NS�}&��T� a � ��}�_ d��Ə֏�0�`��,a aA,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� ܇BpAprbB� � #�`p*�}�%� D�p5�A8<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� ��4Un� ��4o� �tD�m� ��4�!&��Tum� ��4U�g�!8M�۶�`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�W d��Ə֏�0�`��,��a aA,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� ܇BpApqaB��#�`p*�}�%� D�p5�A8<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�[bB!Q}`!8Me� ��4U]�b�!8MU㶘`NSպ1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ d��Ə֏�0�`��,a `A,@��EzAPF�}@�^�T�xAPJ��^��}��!Ȉ�� }�E�9� \�\��}�` ��'��l�@<�`NSټu&ұX`� �I&��T�m� ��4Սe�!8Mu�f�`NSݼ]&��Two��`NS�i&��T�n� ��`4n� ��4�m� �tD�m� ��4Un� ��4�!&��Tum� ��4U��aB!8M��f�!8M��ƙ`NS��#��#��#�&�햙`��5&�0XeBý�aB4CM�b�!8Me�Ƙ`NSٸ9&��Tvn� ��4��[d!�a A ��}�X d��Ə֏�0�`��,��a `A,@��EzAPF�}@�^�T�xAPJ��^��@@��}��!Ĉ�� }�E�9� \�\@�@��}�` ��'��l�@<6�`NSټu&ұX`� �Q&��T7m� ��4�m�e�!8Mu߆�`NSݹ�#@3�f�!8M�Ӷ�`NS�q&��T8n� ��`4n� ��4�m� �,��aB!8M��aB!8M�`!8Me� ��4U][bB!Q}�b�!8MU릘`NSո1&��T5o� ��4U��>B?BX?B Xa!�n� 0V�P�p�v�P�P��`NSٴA&��T�m� �,�I&��T�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p+�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��V� 8�&@3�6�PNS�V�PNS妘PHGT�v�PNS��@NSYb�!8MU׶�`NSո}&��Tym� ��4��>B?BX?B �a!�n� �,��h�!8M��0�`� ��4UޛeBýbB4CMc�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a � ��}�_ 7x`��@@� d �0�c�`�A`��`�0�@N3�?!�a yE,@��P#UzAP��}@!�^��xAP��^��}��!Ȉ�� BpAprbB� � #�`p(�}�%�pCPh0��,�0P�~X��pC�h0�0�,1P��00�� D�p5�A8H�`NSټ�&$@#��&ұ�`� �y&$@#��M&��T�m� ��4Ս[e�!8Mu�֙0�c� ��4��f�!8Mu��P̓��0�`� ��4^�f�!8M��F�@̀�`B!8Me[aB!8M��bB!Q}�aB!8M�[`!8Me� ��4U]�b�!8MU��`NS�q&��T8o�`�A`��`� �tD�i&�,xn� ��4U��c��&��Tyo� 0�n� �5m� �c�!8Me�֘`NSٸ=&��Tvn� ��4��d!�a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� ~ �p1��Q�D8��0�c�`�A`��&��T��#�,>b� ��4a �� p1��Q�D8 ��0�c�`�A`��&��T�&��T�@��a �� G� �p1��Q�D8��0�c�`�A`��&��T�&��T�`��a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� gpq0l@`A�\�7�\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}e�!8Mu�&�0�c� ��4��d�!8Mu��`��ؼ&@3�e�!8M�Ӗ�`NS�a&��T�m� ��`B!8Me[aB!8M�a4n� ��4o� ��4��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� gpq0l@`A@\@7�\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�a!�n��ÿ8v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a bA,@��UzAP��}@!�^��xApr2b` ��BpAp r2b` ��BpAp r 2� ipq0l@`A�\�7A\@\@\��l�@<��`NSٽ]&��q&ұ�`� �A&��Twm� ��4�}�d� 0�U&��T7o� ��4Սe�!8Mu��`��ؼ&@3�e�!8M�ӆ�`NS�e&��T�m� �a4n� ��4�m� ��4Un� ��4�&��T��#OD��b�!8MU㖘`NSպi&��T8o�`�A`��ÿ8f�@HG��f�!8M�Ӷ�`NSٴ1&��T�m� �,�5&��T6a bA,@��UzAP��}@!�^��xApq1b` ��BpAp q1b` ��BpAp q 1� ipq0l@`A@\@7A\@\@\��l�@<�`NSٽq&ұ�`� �M&��T7m� ��4�][e�!8Mu�f�`NSݺI&�f�!8Mu��`��ظe&��Two� ��`NS�m&��T8n� �a4n� ��4�m� ��4Un� ��4o� ��4�!&��T5m� ��4Um[?��D��-&��T�n� ��4U��>B?B�?��D��/�&��v�P�s�֘`NSٴ)&�,��`NSٹA&��T�a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �G��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��\Ə@��a ��A� ��\Ə@��a ��A� �H��Ə@��a ��A� �H�� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a ��A� �� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H�� #�,a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �� #�,a ��A� �H�� #�,a ��A� �ȿ�� #�,a �� A�@��Ə֏@��a ��A� �Ⱦ�� #�,a ��A� �H�� #�,a ��A� �Ƚ��Ə@��a ��A� �G��Ə@��a �� A�@��Ə֏@��a �� Ј� A�@��Ə֏@��a ��A� �Ȼ��Ə@��a ��A� �H��Ə@��a ��A� �Ⱥ��Ə@��a ��A� �H��Ə@��a ��A� �ȹ��Ə@��a ��A� �H��Ə@��a ��A� �ȸ��Ə@��a �� A�@��Ə֏@��a ��A� �ȯ��Ə@��a ��A� �H��Ə@��a ��A� �Ȯ��Ə@��a ��A� �H��Ə@��a ��A� �ȭ��Ə@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏0@�c��,a ��A� �H��Ə@��a ��A� �Ȫ��Ə@��a ��A� �H��Ə@��a ��A� �ȩ��Ə@��a ��A� �H��Ə@��a ��A� �Ȩ��Ə@��a ��A� �H��Ə@��a ��A� �ȟ��Ə@��a �� A�@��Ə֏@��a �� H� B�A8��Ə֏�@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a ��A� �H��Ə@��a ��A� �Ȝ��Ə@��a ��A� �H��Ə@��a ��A� �ț��Ə@��a ��A� �H��Ə@��a �� H� B��@��Ə@H�Y?!<Sa �� Ț B��@��Ə@H�Y?!<Sa ��A� �ș��Ə@��a ��A� �H��Ə@��a �� A�@Ə��4֏@��a ��A� �H��Ə@��a ��A� �ȏ��Ə@��a �� A�@Ə��4֏@��a ��A� �Ȏ��Ə@��a ��A� �H��Ə@��a ��A� �ȍ��Ə@��a ��A� �H��Ə@��a ��A� �Ȍ��Ə@��a ��A� �H��Ə@��a ��A� �ȋ��Ə@��a �� ɋ A�@��Ə֏@NS�?!�a � �� ɋB�@��Ə֏@NS�?!�a �� A�@��Ə֏@��a �� A�@��Ə֏@��a "D,��}��\`��#�``&-v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&+v��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� A�@��Ə֏@��a �� I A�@��Ə֏@��a � �� B�@��Ə֏@NS�?!�a �� A�@��Ə֏@NS�?!�a �� ܇~ A�@��Ə֏@��a �� ܇~ A�@��Ə֏@��a "D,��}ğ\`��#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa "D,��}��\`@�#�``&�u��@�A��A8&�`��Է�#��T�&��T��#��T��Ə�@��`BS,�Oa �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� } A�@��Ə֏@��a �� A�@��Ə֏@��a �� z A�@��Ə֏@��a � �� {B�@��Ə֏@NS�?!�a �� { A�@��Ə֏@NS�?!�a �� y A�@Ə0T�b�0!I֏@��a �� Iy A�@Ə0T�b�0!I֏@��a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a ��A#�`�&�5�Ə0T�b�0!I֏04�a��,a �� Ix A�@Ə0T�b�0!I֏@��a �� o A�@Ə0T�b�0!I֏@��a �� Io A�@Ə0T�b�0!I֏@��a �� n A�@Ə0T�b�0!I֏@��a �� In A�@Ə0T�b�0!I֏@��a �� m A�@Ə0T�b�0!I֏@��a �� Im A�@Ə0T�b�0!I֏@��a �� l A�@Ə0T�b�0!I֏@��a �� Il A�@Ə0T�b�0!I֏@��a �� k A�@Ə0T�b�0!I֏@��a �� Ik A�@Ə0T�b�0!I֏@��a �� j A�@Ə0T�b�0!I֏@��a �� Ij A�@Ə0T�b�0!I֏@��a �� i A�@Ə0T�b�0!I֏@��a �� Ii A�@Ə0T�b�0!I֏@��a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� ,�5A�@֏0T�b�0!IƏp�a��,a �� _ A�@Ə0T�b�0!I֏@��a �� I_ A�@Ə0T�b�0!I֏@��a �� _ B�A8֏0T�b�0!IƏp�a��,a �ȈA� �I^� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a �ȈA� ��]� ��@�@ �0�c��4��#��T��PNS��PNS�֏@��a ;J,@ �� a��pC�`0�0�� ^%�1K�7�2F0bp v�W��#�``'|�,�,�1P�D!��@�z(�p�$��A8F�`��Խ&��`B!0OM[b�!8̀Ӧ�`N3�&�,7m��&�,wo� ��<vm� ��4��#��#��#� &�,a 9J,@ �� ЕQ�$� E��,�@#�` &u��p�Q��,Ca#�` &v�$�ǈ�� ]-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 9J,@ �� Q�$� E��,�@#�` &p��p�Q��,Ca#�` &q�$�ǈ�� I\-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3�@��a!8�`� ��<�&�,vm� ��<��Əf�`��ػ&�Թ%&��vo� ��4��a 9J,@ �� Q�$� E��,�@#�` &k��p�Q��,Ca#�` &l�$�ǈ�� [-�,�1P�D!��@�z(�p�$��A86�`��Է�#�3v�@N3�`!0�n� �,�]`� 0��#��#�&�,�n� ��<un� ��4��[b�!8�`��@��a 5J,@� ��,C #�`"cE�$� ��,�@#�`"de� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`">E�$� ��,�@#�`"?e� � F��A��2Kp��A��U2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'9E�$� ��,�@#�`�':e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 5J,@� ��,C #�`�'4E�$� ��,�@#�`�'5e� � F��2Kp��T2KpTD!��@��1K�TP ��0�c�`� ��4��#�3f�`N3Զ&��m� �,Kn� ��<� &�,� &��@��a�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP �"S�,�1bP �"S�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a 7J,@ �� T1K�7x��1K�7�2F0bP ��R�,�1bP ��R�,�1P�D!��@�z(�p�$��A8��0�c�`� ��4��#�3v�`N3Զ!&��m� �,Kn� ��<�&�,� &��@��Xa�!�R�a � �X�@��Ə@0�@X?!8Ma � �T�@��Ə@0�@X?!8Ma � �`�@��Ə@0�@X?!8Ma �� H A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� ? A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� > A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� = A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� < A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� ; A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� : A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 9 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� 8 A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� / A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� . A�@Ə0T�b�0!I֏@��a �� - A�@Ə0T�b�0!I֏@��a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a ��A#�`�&�4�Ə0T�b�0!I֏04�a��,a �� , A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� + A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� * A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ) A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� ( A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� ,94A�@֏0T�b�0!IƏp�a��,a �� ,74A�@֏0T�b�0!IƏp�a��,a �� ,54A�@֏0T�b�0!IƏp�a��,a �� ,34A�@֏0T�b�0!IƏp�a��,a �� I B�A8֏0T�b�0!IƏp�a��,a �� B�A8֏0T�b�0!IƏp�a��,a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �� A�@Ə0T�b�0!I֏@��a �ȈA� � ��Ə@NS�>�D�`��,a �ȈA� ȉ��Ə@NS�>�D�`��,a �ȈA� � �� @�@ �0�c��4��#��T�&��T��#OD֏@��a �ȈA� ȉ�� @�@ �0�c��4��#��T�&��T��#OD֏@��a ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa ��A� �J��0<Sa ��A� ��0<Sa �b�d�@��<a �a�d�@��<a ��d�@��<a ��d�@��<a �b�p�@��pD� �a �a�p�@��pD� �a ��p�@��pD� �a ��p�@��pD� �a A, �\`d�`�n@��`tdld{@�;�;#�`�*��A8��0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#�`P��A8��pD� �a �� G�B�A8֏`8E!��# )Ə@�Da ��A� �J�� a �� Ə0�d�2a ��A� �J��@0��a �pCPa0bp ��7�� Ap�A8Ə@<1�>��a A,@ � �� ؊�B0��Ə@<1�>��a �pCPa0bp ��7C� Ap�A8Ə@<1�>��a A,@� �� ؊�B0��Ə@<1�>��a A, 4�\W ��8�C:�C9�C8��:��8́��} #�` a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұX?�!EB 4��#Y`� Q��# )a A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#֏pH�P a� �� &ұ��#�>��a A,@�&!P�܈A� � � �a� �A8 �PX�0&�0H�b�CD֏P�2��@0��X`� Qa A,@�&!P�܈A� � � �a� �A8�PX�0�PHT�#&�PH��#Ə0T@d��@,`�H a A,@�&!P�\Î$��M��&@�@�PX�06� 8� &ұ?�P��#�L��# )�0HTa A,@�&!P�\Î$��M��&@�@�PX�0�PHT�#6�@8��`B!��?�P��#�L��# )a A,@�O6!P\Ď$��B$H0��PX�0F� 8� &ұ?�P��# )&�0HT`��4� H a A,@�O6!P\Ď$��B$H0�� PX�0&�PHT�#F�@8��`B!��?�P��# )֏�,@C��a #A,@�&!P�\Î�� 2b� ��7J�0�#�`�+rd�PX�06� 8� &ұX?�!EB 4�&�,�m�CD�0HT`�H�&�,a A,@�&!P�܈A� ��# �`�+y��!�&��A� �J�� PX�0f�`��Է &ұ?�P��#�L��# )�0HT`� �,a "A,@�&!P�\Î�� 2b� ��7 �0�#�`�+rd�PX�06� 8� &ұ�a�!�R�Ə0T@d��@,`�H�&a!�a #A,@�O6!P\Ď��M2b� ��7K$�hB��M�A8�PX�0F� 8� &ұb�!�R�Ə0T@d� �D��@0��X?� a�b� �,a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H��#R$�@C�b�!Q��Ԇ�@8��aBHLDH�2?�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�&ұ��a�HLDH�2��-& �H�#��`8��a FA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�PHT�#��`H��#�&(��D�D-Sm� �t,>b�H�)&�Hm� 1!Q�d�CD֏P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�PHT�#֘P8��Ƙ`H��#�&(��D�D-Sm� �t,>b�H�-&�Hm� 1!Q�d� ��#Y?B1�a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a JA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�PHT�#�P8��֘`H��#�&(��D�D-Sm� �t,>b�H��# �& )Ƙ`HT�#�%&N�#V��LƏ0T@a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a MA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#��#R$�@C�b"Q��Hm� �t,>�#�&$��D�D-��#�a�!��a GA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0V��L�&��>��b� �D>�#�-& ұ��F��LƏ0T@d��@,`� �t,>�#a LA,@�5WP��`tdl�@q�d ��F��7�P��`t�@p�d��`@r�G��<�9#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`@*O`d�PX�0V��L�&��>��b� ��Xc��D>�#�-& ��Ƙ�H��#>baBHLDH�2?�P��#�L�&ұ��a MA,@�OPQsu��`n v@`@p�`�dl@r�{`�9��<��<�{sy��`n v@`�dl@q�ld {p�<�A��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0f��L�&��>��b��4� H�%&N�#>b� ��Ƙ�HT�#>R[c"��m� 1!Q�d�CD��pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0֏pH�P a� �D>�#֘p8��>b� �t,>b��&0��D�D-S�#Ə0T@d�H�!&N�#>b� �t,>�#F��L>b� �D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:�)(Q0�#�`�>N`d�PX�0��HT�#��#R$�@C�aB"��aBHLDH�2��b�D>�#>b�CD��@0��aB"Q��XcB�D>r��c�"��>�#V��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0f�pHT�#>b� �t,>b��&0��D�D-S�#Ə0T@d��@,`�H�&ұ��a�HLDH�2��)& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��$��M�K0��0!P(܈A� �J�D�� 0b`` ��E�@�PX�0��HT�#��&$ұ��&4��D�D-��#Y?B1��>��b� �D>�#>b� �t,>b��Xa�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0f�pHT�#>b� ��#Ƙ�H��#��XaHLDH�2�>b�CD֏P�2��@0��b��D>�#F��L>b� �t,>�#��HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;{su��`n v@`@p�`�dl@r�{`�;��<��<#�`�>�1�� B`��U��A� ��8�@0�hBӄ@�p4;b� �2:S)(Q0�#�`�>N`d�PX�0��HT�#��&$ұ��&4��D�D-��!& N�#>�#Ə0T@d��@,`�H�&$��5&$N�#��1&(ұ��#>b� 1!Q�T��a MA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0v�pHT�#>b� ��#֘�H��#��aHLDH�2�>b�CD��@0��X?� a�b� ��b�!��&,��D�D-��c"Q��>a PA,@�O�zsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�>2�� B`H��U��A� ��< �� `��U��A� � �P�"� 0b`` ��F�@�PX�0Ƙ�HT�#��!&$ұ��&4��D�D-��%& N�#>�#Ə0T@d�H��# �& )v��HT�#>�#昐8��>�#֘�H��#��&8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pn0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��pnPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� �_�O�H#�d� ��4UMaB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�&��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &��T&��L��#�Y`�HLDH�2�F�`NSո&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8��0<Sa 1A,��7�a�� n0�� n0x�@p�d��`�d ��G��<�;#�`�>�0�� Bpp0�p�lC ��L��L��#��`�!8MU��@NS�`B��>�HD4aB��>�HD4a KA,t��7�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L� &(��D�D-o� 1!Q˴a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,D��7b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��7h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��7b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,t��7h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8��0<Sd��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8��0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@��0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8��0<Sa 1A,�� 8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a KA,��8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a .A,t�8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,4�!8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��)8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��18h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��98b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8��0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8��0<Sa .A,��E8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a AA,��Q8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa .A,��]8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6�@NSY`�HLDH�2�&��L�&(��D�D-�n� ��4U�[a�!8MU��L��0<Sd� �_�O�H#�d� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2�6��4MV��4M��L��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#�Y`�HLDH�2�6��4M&��L�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,��7Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`�>�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,t��7W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�` ?�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,D��7�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`P?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��7�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��7�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,t��7�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�?�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��8�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�?�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8��0<Sd��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8��0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#�`B�@��0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8��0<Sa 5A,�� 8Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`B�0�� Bpp0�p�lC ��L��L��#�Y`!8Me� ��4UM�`B��>�HD4aB��>�HD4a OA,��8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`0B�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��L�&(��D�D-�n� 1!Q��a�!8MU㶏0<Sd� ��4U��?BHLDH�2�aB��>�HD4�`!8Me� �_�o�H#�d� ��4�m�bB��>�HD4bB��>�HD4a 2A,t�8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`PB�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,4�!8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`pB�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��)8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��18�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��98�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`�B�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8��0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8��0<Sa 2A,��E8�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a EA,��Q8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�`0C�0�� B0�0� �0�p�lC@� !�6ĐA8��L��#��`�HLDH�2��L� &4~!>�#�DD�&4~!��#�DD��#$��D�D-�&4~!~�#�DD�&4~!��#�DDa 2A,��]8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�``C�0�� B0�P�lC ��L��L��#�Y`B��>�HD4�`B��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #�`�C�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G # �`�C�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`�CC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,�\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,4�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#�` FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,T�\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`@F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#�``FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#�`�C`8��0<Sd��4��#�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8��0<Sd�E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8��0<Sd�E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8��0<Sd�E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#�`�F�!��0U�B~!~��d� ��4n� ��4��#��?�TE��O\S��#�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG# �`�F�BB��@�0U�B~!��d� S,��>qM L6�PNS�F�PNS�&�@NSYa�!8MUӶ�0<Sd�E,a�S,��>qM L��0U�B~!~��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�FC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`�F�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#�`GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,4�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �` G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,T�\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#�`@GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8��0<Sd�E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8��0<Sd�E,a ,A,t�\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#�``GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�G�AB��@�0U�B~!��d��3EƏPT�B�0U�B~!>��d�S,��>qM L�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#�`�GC!��0<Sd�S,��>qM L��0U�B~!~��d�E,a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�Ə ��#�c��t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�Ə ��#�c��t �a �� /A�@ƏPT�B��@0��a �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@��@0��?BQEa �� /A�@ƏPT�B��@0��a �� /A�@ƏPT�B��@0��a ��A� ��-#�`p�0��@0��X?!�b�E,a ��A� �I-#�`;�0��@0��X?!�b�E,a �� -�hB�&�İQ�C8֏@<1�>��b�E,a �� -�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a �� N,�hB�&�İQ�C8ƏPT�B֏@<1�>��a �� N,�hB�&�İQ�C8֏@<1�>��b�E,a ��A� �.#�`�;�@� 0� ��@�@֏@<1�>��b�E,a��,a ��A� ��-#�`�;�@� 0� ��@�@ƏPT�B֏@<1�>��b��,a A, $�\U�C8��;�;�C9��A #�`+��@�� 4a �� O*B�A8֏0T�b�0!IƏp�a��,a �� ܇)A�@��@0��X?!�b�E,a �� )A�@��@0��X?!�b�E,a �� ܇(A�@��@0��X?!�b�E,a �� (A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��?BQEX?!�a �� ܇A�@��@0��X?!�b�E,a �� A�@��@0��X?!�b�E,a �� ܇A�@��@0��X?!�b�E,q y2"�^�4�uJ�`�4�L�b�48�O�d�5Xf[�f�5xv^�h�5��h�j�6��k�l�6�fo�n�7�vz�p�7�|�r�78��t�8Xg��v�8xw��x�8�ǘ�z�9��|�9�g��~�:�w��:`��:8p��;X��;x��;�`Ƀ��<�p̃��<�`σ��=�ڃ��>8�郔�>X��?x��@�0��@��A��B��.��C�?��D8�O��EX_��FxPn��G�p}��H��I�@��J�`��KP��K8ʄ��MX�ބ��N��턺�N��N��OH��O��O��O��O8��Ph� ��P��P��P(@��Ph��Q�`��Q��Q(��Qh��Q��R�)��R((�*��Rh(,��R�(�-��R�(0/��R()�8��Sh)0:��S�)�;��S�)@=��S(*�>��Sh*@H��T�*�I��T�*�K��T(+ M��Th+�N��T�+ X��U�+�Y��U(,P[��Uh,�\��U�,p^��U�,�_��V(-�i��Vh-0k��V�-�l��V�-@n��V(.�o��Wh.Py��W�.�z��W�.P|��W(/0~��WH/��X�/P��X�/И��Y�/p��Y�/��Y8��YH8��Zh80��Z�8��Z�80��[�8л��[9p��[89��\X9�ȅ��\�9�ʅ��\�9�̅��\�9�΅��\8:�؅��]x:�څ��]�:p܅��]�:Pޅ��]8;0腵�^x;ꅹ�^�;셽�^�;��^8<��_x<��_=P ��`�=��aX>�(��b�>p/��c�?�>��c�?H��d�?�I��d(H�K��dhH�M��d�H`O��e�H@Y��e(I [��ehI]��e�I�^��e�I�h��f(J�j��fhJ�l��f�J�n��f�J�x��g�J�z��g(K�|��ghK`~��g�K@��h�K ��h(L��hhL��h�L��i�L��i(M��ihM`��i�M@��j�M ��j(N��jhN��j�N ��k�N@��k(O@��khO ��k�O��k�O�Ȇ��l(X�ʆ��lhX�̆��l�X�Ά��l�X`؆��m(Y@چ��mhY ܆��m�Yކ��m�Y�߆��n(Z�醤�nhZ�놨�n�Z�n�Z`�o([��oH[��oh[��o�[@��o�[��o�[��p\P ��pH\��px\� ��p�\��q�\��q�\`��r]�)��r(]p-��rh]�9��s�]>��t�]M��u�]�^��v�]�o��w^��x(^p��zH^ଇ��{h^��|�^�؇��}�^0��~�^@��^��̀_��(_��H_@��h_@-��_p<��_�K��_`X��_�[��օh�h��ֆ(h`x��ɇHh��hh��h ��҉�hP��Ŋ�h��hp��i��(i0��Hi0��hi`��i��iP��Џ�i��i��jP��(j�*��Hj�9��hj=��j@L��j�X��j`\��j�_��k k��(k {��HkP��՘hk��ș�k@��k��kp��k@��lp��(l��Hl��hl0��l��ٟ�lP��l� ��Ơ�l��m��(m@-��Hmp<��hm0I��m�L��m`Y��m0i��֦�m`x��ɧn��(n��ߨHn ��ҩhn��n��Ϊ�n��np��ޫ�n��o��ޭ(oP��ٮHo��ho@��ү�o��ڰ�o��ձ�o�(��в�o@8��x<��(x�K��Hx�X��hx�\��x�h��x@l��x�|��xp��y ��(y`��Hy��hy`��y��y`��y��y��z��(z� ��Hz��hz��z��z`.��z=��z�K��{�X��({�\��H{�i��h{y��{��{p��{p��{ ��|��(|��H| ��h| ��|��|P��|��|��}��(} ��H}P��h}+��}�9��}�=��}`L��}@Y��~P]��(~`i��H~`m��h~�|��~��~@��~0��~0��P��(��H@��h��(��H� (��h��.��=��pJ��ȈpN��舐[��j��(��y��H�0��h�0��П��ȉ��Ю��(�p��H��x��@��(��H��ȋ`�� H��]��<threadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_89_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64llvm.nvvm.associate.access.property.p0i8.p0i8llvm.nvvm.memcpy.async.p3i8.p1i8__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relumax__vimax_s16x2_relu__vimin_s32_relumin__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32_Z3maxjj__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32_Z3minjj__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfdumin_Z3minij_Z3minji_Z3minllllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindfumax_Z3maxij_Z3maxji_Z3maxllllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS_Z17__reduce_add_syncjj__reduce_add_sync_unsigned_impl_Z17__reduce_min_syncjj__reduce_min_sync_unsigned_impl_Z17__reduce_max_syncjj__reduce_max_sync_unsigned_impl_Z17__reduce_add_syncji__reduce_add_sync_signed_impl_Z17__reduce_min_syncji__reduce_min_sync_signed_impl_Z17__reduce_max_syncji__reduce_max_sync_signed_impl_Z17__reduce_and_syncjj__reduce_and_sync_unsigned_impl_Z16__reduce_or_syncjj__reduce_or_sync_unsigned_impl_Z17__reduce_xor_syncjj__reduce_xor_sync_unsigned_impl7.0.1nvptx64-nvidia-gpulibsmoduleOutputBC��5b0$IY��~-D2!G|!�#�A�I29��%�b�(EB�BD28K 2��H� CF��2BI�%�PAQ��QFQl� ��ap{��vp�ppwxzzH�wp�y�z��pp�y�vw��r `C��?mH�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9�!�6�,@�� :�� ށ��a~��!�a��@�l X6�,�$� �b��X�@0 �f� �3b��x�@@ �� e ��H5��=�C;�A8��;�=�=��;��<�=�C8��<�;��;�C9��5�� !�!��A��!܁ʁ� c@��H�;؃=��;��;�C8�C:��;��/Ѓ<��;�<�C8�=�9� �`a��HI�`� �A&�1ad�$�@�0�2af$K3�ƙP��<�@ah�DɄ��H�iB@M�h�T�`� �"�gB`M*�� 2"� d�%��%ㄡ�L��DM��#M%L>�4Ӎ`!n��&@� � �"�("��4E�0��|�M�O@d�>�,5?�AL�@HB�<�4 D@�(�P�( %�$�#@��"TYp�4E�0��+M0�4��eP4�6(PISD �ϹDE4�/Mc}P2 �@(#@$�2 @(H��2i�(a�j>��i�� U`,P@Pe�� "P�@�� @@dCI�� 0�2 ��&_�@�"#H�s�)��H `�U�Q��$ĐN"*0�.��&_��&��`�&M%L�@ �I0�] D��x�az� (��Q *��h�rP�`@TD)�2�#J�`%Q��%�P�P��(T0�d@Q�� 2�)�P�T�@��#JqE�mQ�+*��-��2TP�� 5"PA1� 1��8F�"J�|�� Ȍk�)��ɧ��jPP4(�4��&_8��F�a ( 6��&� ��@ep� �c�)�� @" (��F�"J�|J8��FPP:(��i�(a��pDP@�� (��i�(a�)�"P@�Ѓ (��#��&_:�h�Q�G��8G�"J�|J:�h�Q�G��F�"J�|�p��FB}��>��BΑ��&��'�l$��G�B�i�(a�É� }��G!飏>h�s�)��ɧ�É� }��G!飏>ȑk�)��ɗ'�pD��G}��>��i�(a�)�p� GD}��G)&飏>� Jn��&i��c�)��ɧ�f�Q@2�4�E�"J�|��l$@8��i�(a�)��l$@>�i�(a�f��B�F)D�r�4E�0�H3�(�)��1�Q��SH3�(�*��1�Q��/Nc�(0��@�\#M%L>�8�Q��"�b�r�4E�0��60��`�r�4E�0��4��(�-@�#M%L��4ƍ0��`��5�Q��S��70��B�b�P/�HSD ��8��#�0��`�0�HSD �O-Nc��0�(�0`�4��5�Q��L5 �)�1��1�HSD �_,0�H(��2$SL1��5�Q��L7�)��B0SL1�P3�HSD �_,0�(��b�)�3�SL1�A�Q��/Nc,Nc�(0��@�\$M%L>�8��8�Q��"TcYs�4E�0��60��`es�4E�0��4��4��(�6@�$M%L��4��4ƍ0��`��E�Q��S��70��B�c��8ISD ��8��8��#�0��`�P9ISD �O-Nc,Nc��0�(�9`��9�Q��,D�u��:��9�Q��,��u�)�;�C�#M%L��P�B�hPGu �:�C�#M%L��P�B��PGu�)=ꨣ�:TO��O!� h�RT@��b�"�T��PL1TP�Q�*C1�C ��b��ʠ�:PCTL�)�!��IT�b�1�(*D1��YT�!� QL1�C�bz�8*C1C�Au�!�ʠ�:�HeP�u��2`�dRPB��<T�T0�X*� `�T ( ��$S)�QGu � ��:�C6Bu��p*�@:��u ��`�:�S�1��@B�A �ʠN1�Q�N�AJJ�P@KA� ��"P=U�+��p�TF@�BTZ�U��*FU�V�(�b��*B1�W�0��*�:�$Vh��*�Y�G�@�Ң5��Z��Q�;ܢ��"�E��.��Ex�)E8E/ڋ�)6}�_�F��#=��`DR�C��h�"L�)F��ƨ1R��<F��6E�F2��<D�A��2J��dF��3Ex�3��)�;�he��H�2\QDZ�@�2�W(�V�+�Z�T+�n�A��ኸ!��p]�le��.NH�2\��)D[��W��V�+Ӄt+��B��&!��pc�pe��1UH�2\!�6D\�x�2�W �!W�+2K��+��=C̕��!��te�bH�2�:D]!Ⱥ2T�]!�H�2T��]!�ȻB�: x��>� ��PAye��ȼ"h�W�H�\A�!�̒!��g�|F�2�D��2�t�� E"PE�3��+�@�a>u�"̷ h�"�h�>�(q��裏>�(��tтE��.j��E��a_$a��(,��E8�0° &Fa>ƈ�"�yX�%#�PL1E13��SL1E(�32��S?��pEЊ��b-��q�[�b!��>(�b\AW}��X�+��W�@7⊾0�r,��X�+�q�a}�A?��裏>(�B\�АŸ�1*�WHf�td!�Ȭ�b(�b\�Y1�S-Y�+>SGj,w�x�|�:hw��t �t�68x�vHq�� !CF��!�0�@-JM5퀀$8r2�`� ��0;�� e�p�d0��`�`��q�c0�($r�8�1�`��`�@��\0} ��L.�>��'��v@~��я�@;��` �|�q�c0�(�r�8@2�`$v�`!@�ڿ\0� ��b��1��@;��r�8H2�`d�A.GI�~5��0�` ��`'@��v@��r�d0��Y@� ��x@�ѓ�@;�qd9|2�`T��+GJ�FPp�P�` �P, �`+@��I��%��v@�)��q�d0�hOW0� �� @;��\�8`2�`TƂ+GL��Yp��` �Ph�`3@�zq��&��v@�.��q�d0�h�W0� �� q��@;�S\�8p2�`�T��0D��s�"��A@��s��!R��@��G�L `��f0D��"��A@��u� �H1;|C$��!��Y0D�� H�;xd@��"��C�U<4 `��C$�,5 `�$��@��zp��!�j0D�� H�=hv@�IŇ�"��C�>�P `�D� @��}�F��!R��(0D��* �Hv?p�@�I��"��C�$6��!s� @�f��U `�� @�֒�^��!Rj[,0D"O�� H!JX�@��F l�"�)a�C�:%\ `��D�@�d��z��!��m�@�f��q�`�4�� @�d�x��!R�P:p0D�_b�H!LD�@�I��HP"�1!�C�:&$|�`�t�Ĵ4�t��!�:�Hp0D�g"* �H,MH(@�ɧ i%P"�51�C��&��`�d�E4�ԛhn��!�}�N0D2q�� H4N@`@��̉g,"�9�C$R' � `�D넴�̝x��!R�P\0D*y��H5ODw@��H/"�=!�C��'$� `��4�ܟ�N��!��j0Dz�Bj �HaXL�@��i6"�bA�C�],�� `��E��H��~��!RN�x0Dz��*�H�Y\�@��7��="�g!�C��,&� `�$�U�̴��!��}0DB��H�ZH�@�I]&?"�k��C$�-�� `��|�`��!��N~0D�ۂ��H�[8�@�Iv�(?"�o1�C$�-�� `�t�Ŕ�t�p��!P~0D��H�\@�@�ɖ'?"�s�C$z.�� `�$��d��!RYU~0D��b��H�]T�@�I�f?"�w��C$�.�� `�d��d�p��!�N0Dr��H|^8�@�IՋ�?"�z��C�_/� `��E"@��XB��!��J�0DB��H4_$�@��&D"�}��C��/� `��E"@�L�H��!��0D��" �H�_!@��$D"��C$�/�� `��@�R�HB��!�G�0D�A# �Hbh !@�I �cD"šQ�C$;4� `�ԇ�Q"�H�(J��!R*ǉ0D�E�8 �H�h %@�� D"��Q�C�t4� `�$��Q"@��@J��!HE�0D*II�H-i)@�)& dE"夡�C��4� `��F�"@��8J��!�^D�0D�K#)�H�i%@��5�$E"ݦQ�C$�4 `�d�F#@��8J��!�|D�0D�Oc��H!j2@��D eE"Ũa�C�5� `��FQ"@��8J��!ҚD�0DzS)�Hwj%@��O �D"��Q�C�T5� `��Q"@�h�@J��!�E�0D W)�H�j%@��] %E"�a�C$5� `�T�Q"@�R�@J��!R�E�0D�Y)�H>k%@�il��?"��a�C$�5� `�ԷƱ"@�H�`h��!��0D�])�H�k%@��|��D"��a�C$�5� `�ĿFQ"@�B�HJ��!� F�0D�a#)�H9l%@�重�D"��a�C$76� `�tǆQ"@��PR��!�)F�0Dje)�H�l%@�I��D"��A�C�r6� `��FQ"@��PJ��!?G�0DiC)�H-m%@��&E"��C$�6� `��׆�"@�D�pR��!lJ�0D�m��H�m6@�ɽ��D"ݷa�C��6 `�t�Ɓ#@�\�Xp��!R��0D s#)�Hjn%@� � �D"�Q�C�>7 `�dꆁ#@�X�Pp��!��0D�u�)�H�n(%@��ލ$E"��Q�C$�7 `��#@�^�HR��!�E�0D�y#I�Hfo)@�i퍤D"ѽa�C$�7� `�4��Q"@�R�HJ��!��F�0D�}#)�H�o%@��D"��a�C$�7 `�d�G�#@�^�px��!R�0D"��(�Hjx%@�� D"��1�C$;< `�ԇ��#@�@�0J��!�)C�0DJ��(�H�x%@�I��G"��1�C$^<� `�䎇S"@��J��!�KN�0D��C)�Hfy%@��-��D"�偔C$�<� `��R"@�t�XJ��!RyH�0Dj�#)�H�y%@��?�L"�聈 C�=2 `��r�� !��}0D��Hzz f@�IT��>�"��Q� C$U=3 `�$��a&@�p�@��!��0DB�1�H�z f@� _�L"�쁘 C$�=� `��"&�p� ��!��0D"��Hl{ �@��n�>�"��C$�=�� `��b&�`��!��0D��1�H�{f@��|�L"� C$�=3 `��b&�L�@��!��0D��I�H>| b@�ɌBL�"�� C$9>0 `��ǃ&�R�`��!R-�0D"�A �H�|<h@�韏M"�� C��>4 `��G�&��h��!�j��0D��Q �H�}<h@��M�"�� C$?4 `��ǃ&��`��!R��0DB��A �H�~<h@�i�M�"�� C$?4 `��ǃ&�r�`��!R��0D��A �H�8h@��M�"�� C��?4 `�$�ǃ&�Ba��!R"�0D�A�A �Hh�@C$<D��!)"�@��f `��&��q��!R?"�0DrI�a �H=�@n@��-�M�"�%r� C$�D 9 `�t��'8�\)��!�l"�0DZO� �H� v@�i?�N�"�(�� C�E0< `�4�H"&@��I��!R�"��0D�S$ �H��`x@��W�DL�"�+�� C�xE1 `��H"&@�F��!��"��0D"[$ �Hm�$b@��n�DL�"�-�� C��E� `�ľ�"@��yB��!� #O�0D�a��Hx�@!@� �(D�"�2�C$tF� `��'�hq��!�X#��0D�k�� H~�8~@��O"�7�� C��F? `��'@�ny��!��#�0Du� �H��8�@��GT�"�;� C$}G&�0D2��a�H5�4l@�I,�M"�e° C$�L6 `��I�&@�p3y��!�y&�0D��a�H$�8l@�IF��M"�i� C�7M6 `�ԧI�&@�T5y��!R�&�0D:��I�H��0�@�IdM"�l"� C$�M.4 `��Ʉ&@�h7��!��&�0D��B�H3�Lh@�Ɍ�M"�qB� C$XN.6 `�t�I�&@��9��!�L'��0D"��R�H{�Lj@�ɵ+U"�w"� C$�N�T `��ɔ*@��<��!��'�0Dj��B�H�\h@�IߓM"�|2� C$�O04 `�� &@�`?��!��'�0D��C�H2�Ph@��M"��R� C$YT06 `��&@�@R��!�M*��0DRKS�H��Pj@��6M"��"� C$�T.4 `�T�ʄ&@��T��!��*�0D�U�b�H�Tl@��_M"Ѭ"� C$�U.4 `�Էʄ&@�`W��!��*�0D�_�b�H1�Tl@�ɌM"�"� C$WV.4 `�T�ʄ&@��Y��!�K+�0Dk�b�Hy�Tl@�ɵM"��"� C$�V.4 `��ʄ&@��\��!��+�0DRu�b�H�Tl@��M"��"� C$�W.4 `�T�ʄ&@�P_��!��+�0D��B�H'�Lh@�I�k `��0D�["��2-@��q��H%�HC��\��!�[.�@�I4��M"��B� C��\07 `�T�K�&@�tt��!�.�0D��B�H��Lh@��]�M"��2� C��].6 `�Ķ�&@�@w��!��.�0DB��B�H��Lh@�I��M"��B� C$>^.6 `�4�K�&@��y��!R?/�0Dr��B�Hd�Hh@�ɮ��M"��"� C��^,6 `�4��&@�L|��!Ҏ/�0Db��B�H��Hh@��֗�M"��2� C$~_,6 `�$��&@��~��!�/�0Dr��B�H�Lh@� ��M"� 3� C�0d.6 `��&@�X��!�82�0D�G�B�H+�Lh@��'�M"�%C� C$�d.6 `��L�&@��QV�!�2��0D2Q�Z�H<�Lr@��M Np"�*� C$Ze&8�`�t�$'@���!�2Y��0D [&��Hr�d��iu�)"�.�� C$�e�8 `�T��*@<�z��!3]��0D�e�{�H��|�@�饙0x�Hp�x�@�ɵ��"�7��C$g�Z `�t�W�F�1�W `��C��g�� `��L�@@�ʞ�b��!�3V�0Dr}F��H��h�@�I�'V"�`s� C$4l�� `��d+@H�V��!�.6�}0DZ�f��H��DC��l&\ `�$�ͥ+@H��p��!k6�0D��H��`�@� FW"�is� C�?m.\ `�d��+�ҵq��!ҽ6��0D*��H;�`�@�in��W"�n3�C��m&p `��.�L��!7�0D��H��\�@�i��\"�s3�C��n,q `�4� &.��!�l7�0D*�f�H��X�@��\"�y��C$>o0p `�D��.�ֽ��!��7�0Dj�f!�H8�,��i��\�"�}��C��o��!�7�0Dz�b�H��@�@��H^"��C�t| `��γ/��С��!�(:��@0DrEgI(�H��,C$rtr `��N2�@҉H��!RL:��0DKGR�Hu�@-@��4f"զc�C��t�� `�d��r3�Bԉt��!��:Q�0DSG�Hu�0�@�)T�\"��ØC�\us``�T��.L��a �H��0�@�)e��]"��C��u$y `�Է�T/�^ׁ��!��:о0Dz_��H$�0�� &]�"��@��a��!;R�0D2eG"�H��L'@�i��Ie"��C��v"� `��53��aj��!�h;��0DzmG��H��H=@�i�� l"��ÐC$wr `��c.L��a�� !��;L�00Du��"��ÐC$]w"v `��N�.��ݑ��!��;R�0D{��Hp�@�@� ��_"վØC��w�t�`��0D��!�H%�H!@��d"��3�C$>|&� `��O�2��`��!R<>��0DBɇ��H4�07@� ,Ig"��#�C$�|&� `�t�C.��a��!�{>��00D�χ1&�H'�0��iF `��C.��!�>ѻ0Dj�G��H��H�@�)^_"��C��}* `�d�c.P��a� �!��>@�Iu�\"��#�C$�}$� `��t2�Z��T��!?Q�0D��'��H��\3@�)��f"��C�x~$� `�4ҏ�3�j��!Z?�0D��!�H��0��\�"��äC$�~��!|?�0D��'b�H-�D�@��I^"��#�C$Q | `�$��/��!ҽ?��@0D��I(�H/�0C��r `�t��2��H��!��?��0D��gR�H��D-@�I�f"� t�C$5�� `�d�s3�D�t��!-BR�0D2Gh�H�1�@��$!�\"�$ĘC��s``��.L��b �H� 1�@��5��]"�&�C��$y `��T/�f��!ҘBо0D�S��H� 1��V!&]�"�+�@��b��!�BR�0DRYH"�H<M'@��m�Ie"�.�C�م"� `�D��53��bj��!� C��0D�aH��HeI=@�鎡 l"�2Ĉ C$[�� `�$�C6T��b�� !�>C��P0D2i�A�H2 1b@�ɧ!Im"�5$� C$ц&� `�T��7��!R~C�0DRqȲ�H`1d��!�l�"�9Ġ C$R�� `��Є7��!��C�0DRy��H>I�@� �!It"�>��C$݇� `�d��:��T��!R F�0D��r�Hl1b@� #Fl"�bĐ C$X��`�ċc6T��1b��!=F��0D�HR�H2Il@��,��m"�e4� C$Ռ"� `��QD7��3��!��F�`0Drш1,�H:1h@��L#Fl"�i4� C�R�&� `��7��5��!�F�0D��H�Hda�@��o#�t"�nĠC$ڍ&� `�t��:��7�\��!�G��0D��Hf1d@�I�#�l�"�qĘ C$U�� `�d�#6��9��!=G�0D2�hr�H9Mp@��Hn"�u� C$׎,� `�D�C6X��;b��!G�0DB��H4Mx@��̣Io"�yD� C$W�(� `�$�:��=�D��!��G�0D��1�Hj1�@��Iu"�~4�C$ޏ(� `��#6��?b��! J�P0D�A�!*�H`(1f@�� %m"�Ĉ C$R�$� `�d��6��Q��!�>J�0DrI)��H?)Et@� .%�n"��Đ C$ٔ��`��6��Sb��!�}J�0DJQi��H<*Q|@� N%�o"��$�C$[�$� `��F:�@VbL��!�J�0D�YiR�Hk+M�@��o%�u"ɮd�C$�� `��jX��!K�}0De�� $�H�,Y�@��%v"��C$��,� `��ג;�d[�`��!zK��0D�o�� H-.Y�@�i�%�O�"ݹ��C�W�8�@`��f;�F^�t��!��K� 0D�{ ��H�/a�@��%Lw"��C��0� `��+��p��!�+N��p0D��H�8}�@�I&��>"��C$Ԝ�� `��w@��s��!R�N�}0D��H�:u�@��\��>"��C$�� `��w@�`w��!��N�}0D"��H;<}�@�I�'*"��C$s� U�x"��@� �'�|�x"��$q@��'-�x"��d�@��'N `�꓇>@H�|}��!һO�� 0Dr��S�Hl?}�@�i��H}"��I��$�H5Hu�@�I�,V��"�"U�@� ��}"�$��C��X@`�T��>�V�j��!nR��0D�Ojr �H(Jy�@��L)�}�"�)e�C$X�� `�ԫT?��!R�R�0DZ[�� $6��$2�,L�� &G�C`� � �@��! �@!Ȁ(��"(��(�*��(@�D(@�)@�(@@�$(@@��(�R*�B+��+��+ƀ�(ˀ (րr (ـ� �Fh�B�:b`G9�t3��f=�C8�ÌB�yxs�q��3B��Ρf0=�C8��=�C=�=�x�tp{yH�ppzpvx�p ��0n0��P3��!�!�af0�;��;�C9�<��<�;��v`{h7h�rh7��p��p`v(v�vx�w��_�q�r��y��,��0bȡ�̡��a�!ā�a֐C9�C9�C9�C9��8�C8�;��/��<��;�;��i�pX�rp�thx`�t�t��S��P��@� �P3 (��A�!܁��fQ8�C:��;�P$v`{h7`�wxx�QL��P3j�a�!��~��!�aT��8��;�C=�C9��<�C;��;�Ì� �y��w�tz(r��\��P�0#��A��y LbH C�9$��@��h"P2�!GȐQ�align�y0��c0��c0��0��0��0��c0��0�`�0��0� �0��c0�P�0��0��0��0��c0��0�P�0��0��0�@�0��0�!1��$1��%1�`(1��)1�+1�`-1� "v1��c0� "w1� "|1� "}1� "�1� "�1� "�1� "�1��.�5�/�5�0/�5�`/�5��/�5��/�5��/�5� 2�5�P2 6��2!6��2"6��2#6�3$6�@3%6��3(6��3)6�6*6��.26�/36�0/46�`/56��/66��/76��/86� 2=6�P2>6��2?6��2`6��2a6�3b6�@3c6��3f6��3g6�6h6��.p6�/q6�0/r6�`/s6��/t6��/u6��/v6� 2{6�P2|6��2}6��2~6��26�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3�6��3�6�6�6��.�6�/�6�0/�6�`/�6��/�6��/�6��/�6� 2�6�P2�6��2�6��2�6��2�6�3�6�@3�6��3 7��3!7�6"7��.*7�/+7�0/,7�`/-7��/.7��//7��/07� 257�P267��277��287��297�3:7�@3;7��3>7��3?7�6`7��6h7��c0��6i7��c0��6j7�7k7��c0�07l7�P7m7��c0�p7n7��7s7��7t7��7u7��7v7�:w7�0:x7�P:y7��:|7��:}7��:~7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6�7��6�7��6�7�7�7�07�7�P7�7�p7�7��7�7��7�7��7�7��7�7�:�7�0:�7�P:�7��:�7��:�7��:�7��6"� ��#oC��Bܔ`0�07&� �� #w��B޴`0��7.� ��#��B��`0�0:2� ��#�c��B��`0��:8� ��+#��c0��>�� 8#�3��c0�?�� ď9#�s��B��c0�?�� ď:3�0�@�C3Fp�0 �1Ð��3K��043��03P�0HA4�0�D3U�0XA4�p�D3Y�0hA4ð�D3]�0xA4��`0��0�A3c��`0�P�0�A3g��C��C��C��C��C��C��(�C(��(��(�)�C)��)��)�*�C*��*��*�+�C+��+��+�,�C,��,��,�-�C-��-��-�.�C.��.��.�/�C/��/��/�8�C8��8��8�9�C9��9��9�:�C:��:��:�;�C;��;��;�<�C<��<��<�=�C=��=��=�>�C>��>��>�?�C?��?��?�H�CH��H��H�I�CI��I��I�J�CJ��J��J�K�CK��K/1�AL�0�D33��CM41�`�M�0�Dp39��N11ðAL�0�D@3=��O71��M�0�E@3a��X71�0�K�0�E3e��Y41�pM�0�Ep3i��Z/1ðAL�0�E3m��[41��M�0�Ep3q��\71�0�M�0�E�3u��]11�pM�0�E@3y��^71ð�K�0�E3}��_41��M�0�Fp3��h41�0�M�0�Fp3��i11�pAL�0�F@3��j71ð�M�0�F�3��k11��M�0�F@3��l71�0M�0�Fp3��m��mL@3�0M�P�F�41C��nL@3�0M�P�F71C�A��oQp3�D�M�P�G71C!A�� h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�h�n�n�n�n�n�n�n�n�n�n�n�n�n�n�n��职&(7;��49�2��7�Q�H��ή�%��.�.��mn��(�8�H�X�h�x�� )9IYiy��0 12)394I5Y6i7y8�9�:�;�<�=�>�?�P QR)S9TIUYViWyX�Y�Z�[�\�]�^�_�p qr)s9tIuYviwyx�y�z�{�|�}�~�� )�9�I�Y�i�y�� )�9�I�Y�i�y�� r(�w�zXp�C=��8�C9�Â�ơ �A��!�!��<��;�;�=��<�C8��a A,T�� d�\��4Ma A,�� d�\��4Ma A,�� 6C�A8 �\��`&��4M��4Ma A,�� 6C�A8 �\��`&��4M��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,D�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,t�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,�� d�\��4Ma A,�� d�\��4Ma A,D�� 6C�A8 �\��`��4M&��4Ma A,t�� 6C�A8 �\��`��4M&��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,4�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,d�� d�\��4Ma A,�� d�\��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,$�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,t�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,�� d�\��4Ma A,�� d�\��4Ma A,4�� 6C�A8 �\��`��4M&��4Ma A,d�� 6C�A8 �\��`��4M&��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,$�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,d�� d�\��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,$�� d�\��4Ma A,T�� d�\��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,D�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,t�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,�� d�\��4Ma A,�� 6C�A8 �\��`��4M&��4Ma A,�� 6D0�B�@ �\��`�d6��4M&��4MF��4Ma A,D�� A�6E0��lCP�6��4MV��4M�X�\��`�df��4MF��4Ma A,��\ch��n0s n ��dldld {�F��#�`p� d��a A,��\ep��n0q`� n ��dldld {�F��#�`p� d��a A,��\ch��n0s n ��dldld {�F��#�`p� d��a A,��\ep��n0q`� n ��dldld {�F��#�`p� d��a (A,��\��>��<��9��C�(I;�>��<́�=��C8��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I;�>��<ā��=��C8��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I;�>�A=́�=��C8��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I;�>�A=ā��=��C8��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I;��;��<́�=��C:��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I;��;��<ā��=��C:��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I;��;�A=́�=��C:��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I;��;�A=ā��=��C:��C�>��C��A?( ��#�`�1��a ,A,��\��>��<��9��C�(I;�>��<́��<�;��C��A��(I;�>��<́��<�;��C��A?( ��#�`�1��a -A, ��\��>��<��9��C�(I;�>��<ā��<�;��C��A��(I;�>��<ā��<�;��C��A?( ��#�`�1��a ,A,��\��>��<��9��C�(I;��;��<́��<�;��C��A��(I;��;��<́��<�;��C��A?( ��#�`�1��a -A, ��\��>��<��9��C�(I;��;��<ā��<�;��C��A��(I;��;��<ā��<�;��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I8�9��<́�=��C8��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I8�9��<ā��=��C8��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I8�9�A=́�=��C8��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I8�9�A=ā��=��C8��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I8�9��<́�=��C:��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I8�9��<ā��=��C:��C�>��C��A?( ��#�`�1��a (A,��\��>��<��9��C�(I8�9�A=́�=��C:��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I8�9�A=ā��=��C:��C�>��C��A?( ��#�`�1��a *A,��\��>��<��9��C�(I8�9��<́�=��C8��ȁ�C9�C=��=�� } l l l #�`�1��a ,A,��\��>��<��9��C�(I8�9��<ā��=��C8��C�>ȁ�C9�C=��=�� } l l l #�`�1��a *A,��\��>��<��9��C�(I8�9��<́�=��C:��ȁ�C9�C=��=�� } l l l #�`�1��a ,A,��\��>��<��9��C�(I8�9��<ā��=��C:��C�>ȁ�C9�C=��=�� } l l l #�`�1��a RA,5T�\-��>��C9��<�C9��/��C9�<��( ��<�=��C9��<́��/��C9�<��( ��<�;��/��C9�<��( ��<�;��A��/��C9�<��$} l��#�`@?1�&�h�7zP�@�@6��L��V�0�c� ��&��T��&(��D�D-�m� ��4a RA,5T�\-��>��C9��<�C9��/��C9�<��( ��<�=��C9�A=́��/��C9�<��( ��<�;��/��C9�<��( ��<�;��A��/��C9�<��$} l��#�`@?1�&�h�7zP�@�@6��L��V�0�c� ��&��T��&(��D�D-�m� ��4a RA,5T�\-��>��C9��<�C9��/��C9�<��( ��<�=��C9��<́��/��C9�<��( ��<�;��/��C9�<��( ��<�;��A��/��C9�<��$} l��#�`@?1�&�h�7zP�@�@6��L��V�0�c� ��&��T��&(��D�D-�m� ��4a RA,5T�\-��>��C9��<�C9��/��C9�<��( ��<�=��C9�A=́��/��C9�<��( ��<�;��/��C9�<��( ��<�;��A��/��C9�<��$} l��#�`@?1�&�h�7zP�@�@6��L��V�0�c� ��&��T��&(��D�D-�m� ��4a �A,\��\��>��<��9��<�9�<��<��<��9��C��<��<�C��<��<��(I;�>��<ā��;؃��<��<�A?��(I;��=��8́��>��<��<��C9�<�A9ă�C��=��<��<̃�(�<�=��C<��<ā�<��<�C��<��C9�<��8́��A��<��C9�<��8́��A��<��( ��} l��#�`p?AA�&�h��&�p��apq0l@(A7P�@&�@��PNS嶚PNS�V��L��&(��D�D-�m� ��4��&��T5m� ��<��&OD��iBHLDH�2�gB�gBXh��D��/��&�a �A,\��\��>��<��9��<�9�<��<��<��9��C��<��<�C��<��<��(I;�>�A=ā��;؃��<��<�A?��(I;��=��8́��>��<��<��C9�<�A9ă�C��=��<��<̃�(�<�=��C<�A=ā�<��<�C��<��C9�<��8́��A��<��C9�<��8́��A��<��( ��} l��#�`p?AA�&�h��&�p��apq0l@(A7P�@&�@��PNS嶚PNS�V��L��&(��D�D-�m� ��4��&��T5m� ��<��&OD��iBHLDH�2�gB�gBXh��D��/��&�a �A,\��\��>��<��9��<�9�<��<��<��9��C��<��<�C��<��<��(I;��;��<ā��;؃��<��<�A?��(I;��=��8́��>��<��<��C9�<�A9ă�C��=��<��<̃�(�<�=��C<��<ā�<��<�C��<��C9�<��8́��A��<��C9�<��8́��A��<��( ��} l��#�`p?AA�&�h��&�p��apq0l@(A7P�@&�@��PNS嶚PNS�V��L��&(��D�D-�m� ��4��&��T5m� ��<��&OD��iBHLDH�2�gB�gBXh��D��/��&�a �A,\��\��>��<��9��<�9�<��<��<��9��C��<��<�C��<��<��(I;��;�A=ā��;؃��<��<�A?��(I;��=��8́��>��<��<��C9�<�A9ă�C��=��<��<̃�(�<�=��C<�A=ā�<��<�C��<��C9�<��8́��A��<��C9�<��8́��A��<��( ��} l��#�`p?AA�&�h��&�p��apq0l@(A7P�@&�@��PNS嶚PNS�V��L��&(��D�D-�m� ��4��&��T5m� ��<��&OD��iBHLDH�2�gB�gBXh��D��/��&�a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��\��@��a ��A� ��@��a ��A� �� }&�,a ��A� �H�� }&�,a ��A� �� }&�,a ��A� �H�� }&�,a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a �� A�@��@��a �� H� A�@��@��a �� A�@��@��a ��A� �� }&�,a ��A� �� }&�,a �� H� A�@��@��a �� A�@��@��a ��A� �� }&�,a �� A�@��@��a �� H� A�@��@��a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a ��A� �� }&�,a �� H� A�@��@��a ��A� �� }&�,a ��A� �� }&�,a ��A� ��@��a ��A� ��@��a �� A�@��@��a �� H� A�@��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� H� A�@��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� H� A�@��@��a �� Ⱦ A�@��@��a �� H� A�@��@��a �� B�A8��0@�c� �,a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� H� A�@��@��a �� B�A8��@��a �� H� A�@��@��a �� ȯ A�@��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� B��@��@H�h!<Sa �� B��@��@H�h!<Sa ��A� ��@��a ��A� ��@��a �� Ȫ A�@��4�@��a ��A� ��@��a ��A� ��@��a �� H� A�@��4�@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a ��A� ��@��a �� A�@��@NSYh!�a � �� B�@��@NSYh!�a �� ܇� A�@��@��a �� ܇� A�@��@��a �� ܇� A�@��@��a "D,��}ę\`��#�``&rv��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}ԙ\`@�#�``&pv��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}ԙ\`@�#�``&pv��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}ԙ\`@�#�``&pv��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa �� I� A�@��@��a �� ɚ A�@��@��a �� ɚ A�@��@��a �� ɚ A�@��@��a �� ȯ A�@��@��a �� A�@��@��a � �� ɘB�@��@NSYh!�a �� ɘ A�@��@NSYh!�a �� ܇� A�@��@��a �� ܇� A�@��@��a �� ܇� A�@��@��a "D,��}��\`��#�``&>v��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}��\`@�#�``&<v��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}��\`@�#�``&<v��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa "D,��}��\`@�#�``&<v��@�A��A8V�`��Է�&��T��&��T��&��T��6�@��iBS,�Oa �� I� A�@��@��a �� ɍ A�@��@��a �� ɍ A�@��@��a �� ɍ A�@��@��a �� H� A�@��@��a �� A�@��@��a � �� ɋB�@��@NSYh!�a �� ɋ A�@��@NSYh!�a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a ��A#�`�&+6��0T�b� 0!I�04�a� �,a ��A#�`�&)6��0T�b� 0!I�04�a� �,a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� ~ A�@��0T�b� 0!I�@��a �� ~ A�@��0T�b� 0!I�@��a �� } A�@��0T�b� 0!I�@��a �� } A�@��0T�b� 0!I�@��a �� | A�@��0T�b� 0!I�@��a �� | A�@��0T�b� 0!I�@��a �� { A�@��0T�b� 0!I�@��a �� { A�@��0T�b� 0!I�@��a �� z A�@��0T�b� 0!I�@��a �� z A�@��0T�b� 0!I�@��a �� ,�5A�@�0T�b� 0!I��p�a� �,a �� ,�5A�@�0T�b� 0!I��p�a� �,a �� x A�@��0T�b� 0!I�@��a �� x A�@��0T�b� 0!I�@��a �� o B�A8�0T�b� 0!I��p�a� �,a �ȈA� � o� ��@�@ �0�c� ��4�}&��T&�PNS�6�PNS��@��a �ȈA� ȉn� ��@�@ �0�c� ��4�}&��T&�PNS�6�PNS��@��a ;J,@ �� a��pC�`0�0�� ؉o%�1K�7�2F0bp v�[��#�``'��,�,�1P�D!��@�z(�p�$��A8v�`��Խ�&�iB!0OMk�!8̀�֚`N3൥&�,7m� ��&�ص�&�,wo� ��4�y&�}&��&��&�,a 9J,@ �� ܖQ�$� E��,�@#�` &��p�Q��,Ca#�` &��$�ǈ�� In-�,�1P�D!��@�z(�p�$��A8f�`��Է�&�3V�@̳�j!8�`� �,�]�h� 0�y&�}&��&�,�n� ��<un� ��4��i!�b� ��4��a 9J,@ �� ȖQ�$� E��,�@#�` &��p�Q��,Ca#�` &��$�ǈ�� m-�,�1P�D!��@�z(�p�$��A8f�`��Է�&�3V�@̳�j!8�`� �,�]�h� 0�y&�}&��&�,�n� ��<un� ��4��k�!8�`�F�@��a 9J,@ �� Q�$� E��,�@#�` &��p�Q��,Ca#�` &��$�ǈ�� k-�,�1P�D!��@�z(�p�$��A8f�`��Է�&�3��@N3Xi!0�n� ��4��j�!�b�6�0�c� `� A`� �,��iB!0O�i!�b� ��4��a 5J,@� ��,C #�`"�E�$� ��,�@#�`"�e� � F��A��2Kp��A��V2KpTD!��@��1K�TP ��0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��&�@��i�!�R�a 5J,@� ��,C #�`"�E�$� ��,�@#�`"�e� � F��A��2Kp��A��V2KpTD!��@��1K�TP ��0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��&�@��i�!�R�a 5J,@� ��,C #�`�'|E�$� ��,�@#�`�'}e� � F��2Kp��U2KpTD!��@��1K�TP ��0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��&�@��i�!�R�a 5J,@� ��,C #�`�'wE�$� ��,�@#�`�'xe� � F��2Kp��U2KpTD!��@��1K�TP ��0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��&�@��i�!�R�a 7J,@ �� U1K�7x��1K�7�2F0bP �RW�,�1bP �RW�,�1P�D!��@�z(�p�$��A8&�0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��6�@��j�!�R�a 7J,@ �� U1K�7x��1K�7�2F0bP �W�,�1bP �W�,�1P�D!��@�z(�p�$��A8&�0�c� `� ��4�}&�3��`N3Զ�&��m� �,Kn� ��<��&�,��&��6�@��j�!�R�a � �X�@��@0�@h!8Ma � �T�@��@0�@h!8Ma � �`�@��@0�@h!8Ma �� X A�@��0T�b� 0!I�@��a �� IX A�@��0T�b� 0!I�@��a �� O A�@��0T�b� 0!I�@��a �� IO A�@��0T�b� 0!I�@��a �� N A�@��0T�b� 0!I�@��a �� IN A�@��0T�b� 0!I�@��a �� M A�@��0T�b� 0!I�@��a �� IM A�@��0T�b� 0!I�@��a �� L A�@��0T�b� 0!I�@��a �� IL A�@��0T�b� 0!I�@��a �� K A�@��0T�b� 0!I�@��a �� JK A�@��0T�b� 0!I�@��a �� J A�@��0T�b� 0!I�@��a �� IJ A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� II A�@��0T�b� 0!I�@��a �� H A�@��0T�b� 0!I�@��a �� IH A�@��0T�b� 0!I�@��a �� ? A�@��0T�b� 0!I�@��a �� I? A�@��0T�b� 0!I�@��a �� > A�@��0T�b� 0!I�@��a �� J> A�@��0T�b� 0!I�@��a ��A#�`�&,5��0T�b� 0!I�04�a� �,a ��A#�`�&*5��0T�b� 0!I�04�a� �,a ��A#�`�&(5��0T�b� 0!I�04�a� �,a ��A#�`�&&5��0T�b� 0!I�04�a� �,a �� < A�@��0T�b� 0!I�@��a �� I< A�@��0T�b� 0!I�@��a �� ; A�@��0T�b� 0!I�@��a �� I; A�@��0T�b� 0!I�@��a �� : A�@��0T�b� 0!I�@��a �� I: A�@��0T�b� 0!I�@��a �� 9 A�@��0T�b� 0!I�@��a �� I9 A�@��0T�b� 0!I�@��a �� 8 A�@��0T�b� 0!I�@��a �� I8 A�@��0T�b� 0!I�@��a �� / A�@��0T�b� 0!I�@��a �� I/ A�@��0T�b� 0!I�@��a �� . A�@��0T�b� 0!I�@��a �� I. A�@��0T�b� 0!I�@��a �� - A�@��0T�b� 0!I�@��a �� I- A�@��0T�b� 0!I�@��a �� , A�@��0T�b� 0!I�@��a �� I, A�@��0T�b� 0!I�@��a �� + A�@��0T�b� 0!I�@��a �� I+ A�@��0T�b� 0!I�@��a �� * A�@��0T�b� 0!I�@��a �� I* A�@��0T�b� 0!I�@��a �� ) A�@��0T�b� 0!I�@��a �� I) A�@��0T�b� 0!I�@��a �� ( A�@��0T�b� 0!I�@��a �� I( A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� ,|4A�@�0T�b� 0!I��p�a� �,a �� ,z4A�@�0T�b� 0!I��p�a� �,a �� ,x4A�@�0T�b� 0!I��p�a� �,a �� ,v4A�@�0T�b� 0!I��p�a� �,a �� B�A8�0T�b� 0!I��p�a� �,a �� B�A8�0T�b� 0!I��p�a� �,a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� I A�@��0T�b� 0!I�@��a �ȈA� ��@NS�g�D�`� �,a �ȈA� �I��@NS�g�D�`� �,a �ȈA� �� @�@ �0�c� ��4��&��T��&��T�y&OD�@��a �ȈA� �I�� @�@ �0�c� ��4��&��T��&��T�y&OD�@��a ��A� ��0<Sa ��A� � ��0<Sa ��A� ��0<Sa ��A� � ��0<Sa ��A� ��0<Sa �b�d�@�<a �a�d�@�<a ��d�@�<a ��d�@�<a �b�p�@�pD� �a �a�p�@�pD� �a ��p�@�pD� �a ��p�@�pD� �a A, �\`d�`�n@��`tdld{@�;�;#�`�*��A8�0<Sa A,d�\`��=�C8�<��<�C;�A=��}�l�#�`P��A8�pD� �a �� B�A8�`8E!�y& )��@�Da ��A� � �� a �� 0�d� 2a ��A� � ��@0��a �pCPa0bp ��;�� Ap�A8��@<1�g��a A,@ � �� J�B0��@<1�g��a �pCPa0bp ��;C� Ap�A8��@<1�g��a A,@� �� J�B0��@<1�g��a A, 4�\W ��8�C:�C9�C8��:��8́��} #�` a A,@�&!P�\Î$��N��&@�@�PX�0f� 8��&ұh�!EB 4�}&�g��b� �Da A,@�&!P�\Î$��N��&@�@�PX�0F�PHT�#�pH�P a� ��&ұ��}&�g��a A,@�&!P�܈A� �� a� �A8 �PX�0V�0H�b� CD�P�2�@0��i� Qa A,@�&!P�܈A� �� a� �A8�PX�0F�PHT�#V�PH��#��0T@d� �@,`� H a A,@�&!P�\Î$��N��&@�@�PX�0f� 8��&ұ�g�P��&�L�y& )F�0HTa A,@�&!P�\Î$��N��&@�@�PX�0F�PHT�#f�@8��XiB!��g�P��&�L�y& )a A,@�O6!P\Ď$��B$H0��PX�0v� 8��&ұ�g�P�y& )V�0HT`� �4� H a A,@�O6!P\Ď$��B$H0�� PX�0V�PHT�#v�@8��iB!��g�P�y& )��,@C��a #A,@�&!P�\Î��2b� ��;J�0�#�`�+�d�PX�0f� 8��&ұh�!EB 4��&�,�m� CDF�0HT`� H��&�,a A,@�&!P�܈A� ��# �`�+��!�&��A� � �� PX�0��`��Է�&ұ�g�P��&�L�y& )F�0HT`� �,a "A,@�&!P�\Î��2b� ��; �0�#�`�+�d�PX�0f� 8��&ұ�j�!�R��0T@d� �@,`� H��&�i!�a #A,@�O6!P\Ď��N2b� ��;K$�hB��N�A8�PX�0v� 8��&ұ�j�!�R��0T@d� �D�@0��h� a�b� �,a JA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>�)(Q0�#�`�?N`d�PX�0��PHT�#�P8��`H��#��&(��D�D-Sm� �t,>b� H��&R$�@C�k�!Q��Զ�@8��iBHLDH�2�g�Pa MA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>�)(Q0�#�`�?N`d�PX�0�pH�P a� �D>�#�p8��>b� �t,>b��&0��D�D-S�#��0T@d� H��&N�#>b� �t,>�#v��L>b� �D>R�a FA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��$��O�K0��0!P(܈A� � �D�� 0b`` ��E�@�PX�0��PHT�#�`H��#��&(��D�D-Sm� �t,>b� H��&�Hm� 1!Q�d� CD�P�2a HA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��$��O�K0��0!P(܈A� � �D�� 0b`` ��E�@�PX�0��pHT�#>b� �t,>b��&0��D�D-S�#��0T@d� �@,`� H��&ұ��i�HLDH�2��& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>S)(Q0�#�`�?N`d�PX�0��PHT�#��L��&N�#��&ұ��m� �t,>b� H��&�Hm� ��&$��D�D-�}&hB1�a MA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>S)(Q0�#�`�?N`d�PX�0��pHT�#>b� ��#��H��#��jHLDH�2�>b� CD�P�2�@0��j�!��&,��D�D-��k"Q��>b� ��a JA,@�OPsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�?2�� B`H��U��A� ��< �� `��U��A� ��P�"� 0b`` ��F�@�PX�0��PHT�#�P8��`H��#��&(��D�D-Sm� �t,>b� H��& �& )��`HT�#��&N�#��L��0T@a MA,@�OPsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�?2�� B`H��U��A� ��< �� `��U��A� ��P�"� 0b`` ��F�@�PX�0��pHT�#>b� ��#��H��#��XjHLDH�2�>b� CD�@0��h� a�b� ��j�!��&,��D�D-��k"Q��>a MA,@�5WP��`tdl�@q�d ��F��8�P��`t�@p�d��`@r�G��<�9#�`B�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>�)(Q0�#�`@*O`d�PX�0��L��&��g��b� ��l��D>�#��&R$�@C�k"Q��Hm� �t,>�#��&$��D�D-�}&�j�!��a GA,@�5WP��`tdl�@q�d ��F��8�P��`t�@p�d��`@r�G��<�9#�`B�1�� B`��$��O�K0��0!P(܈A� � �D�� 0b`` ��E�@�PX�0��L��&��g��b� �D>�#��& ұ��v��L��0T@d� �@,`� �t,>�#a LA,@�5WP��`tdl�@q�d ��F��8�P��`t�@p�d��`@r�G��<�9#�`B�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>S)(Q0�#�`@*O`d�PX�0��L��&��g��b� ��l��D>�#��& ��H��#>b�iBHLDH�2�g�P��&�L��&ұ��a MA,@�OPQsu��`n v@`@p�`�dl@r�{`�9��<��<P�sy��`n v@`�dl@q�ld {p�<�A��#�`B2�� B`H��U��A� ��< �� `��U��A� ��P�"� 0b`` ��F�@�PX�0��L��&��g��b� �4� H��&N�#>b� ��HT�#>Rl"��m� 1!Q�d� CD��pH��#>a MA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>�)(Q0�#�`�?N`d�PX�0�pH�P a� �D>�#�p8��>b� �t,>b��&0��D�D-S�#��0T@d� H��&N�#>b� �t,>�#v��L>b� �D>R�a PA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>�)(Q0�#�`�?N`d�PX�0暠HT�#��&R$�@C�jB"��iBHLDH�2��j�D>�#>b� CD�@0��XjB"Q��lB�D>r��k�"��>�#��L��a HA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��$��O�K0��0!P(܈A� � �D�� 0b`` ��E�@�PX�0��pHT�#>b� �t,>b��&0��D�D-S�#��0T@d� �@,`� H��&ұ��i�HLDH�2��& �H�#a JA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��$��O�K0��0!P(܈A� � �D�� 0b`` ��E�@�PX�0֚�HT�#��&$ұ��&4��D�D-��}&hB1��g��b� �D>�#>b� �t,>b��j�HLDH�2�>�#a MA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>S)(Q0�#�`�?N`d�PX�0��pHT�#>b� ��#��H��#��jHLDH�2�>b� CD�P�2�@0��j��D>�#v��L>b� �t,>�#暀HT�#��a PA,@��7�P��`t�@p�d��`@r�G��<�;�su��`n v@`@p�`�dl@r�{`�;��<��<#�`�?�1�� B`��U��A� �J�8�@0�hBӄ@�p4;b� �b>S)(Q0�#�`�?N`d�PX�0暠HT�#��&$ұ��&4��D�D-��& N�#>�#��8��>�#��0T@d� �@,`� H��&$��&(ұ��#>b� 1!Q�T��a MA,@�OPsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�?2�� B`H��U��A� ��< �� `��U��A� ��P�"� 0b`` ��F�@�PX�0��pHT�#>b� ��#��H��#��XjHLDH�2�>b� CD�@0��h� a�b� ��j�!��&,��D�D-��k"Q��>a PA,@�OPsy��`n v@`�dl@q�ld {p�<�A��7WP��`tdl�@q�d ��F��#�`�?2�� B`H��U��A� ��< �� `��U��A� ��P�"� 0b`` ��F�@�PX�0��HT�#��&$ұ��&4��D�D-��& N�#>�#��0T@d� H��& �& )��HT�#>�#��8��>�#��H��#��&8��D�D-S�#>a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8�0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8�0<Sa 1A,�� 8�a�� n�� n0x�@p�d��`�d ��G��<�;#�` B�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a KA,t�8�g�� n��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,D�%8b�� n�� n0q``�dl@q�l�@r�{p�:�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,�18h�� n��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��98b�� n�� n0s `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,t�A8h�� n��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��M8b�� n�� n0v@`�dl@q�l�@r�{p�;�A��#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8�0<Sa 1A,��U8�a�� n�� nPx�@p�d��`�d ��G��<�;#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a KA,��]8�g�� n��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,t�e8b�� n�� nPq``�dl@q�l�@r�{p�:�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,4�m8h�� n��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��u8b�� n�� nPs `�dl@q�l�@r�{p�<�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��}8h�� n��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n�� nPv@`�dl@q�l�@r�{p�;�A��#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8�0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8�0<Sa .A,��8b�� n�� n`s `�dl@q�l�@r�{p�9�A��#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��8h�� n��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n�� n`v@`�dl@q�l�@r�{p�9�A��#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8�0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8�0<Sa A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A, $�\`d�@np��pn0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��pn0q``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��pn0s `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��pn0v@`@p�`�d��D��#�`@.��A8�0<Sa 1A,�� 8�a�� n0�� n0x�@p�d��`�d ��G��<�;#�` B�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� �_�O�H#�d� ��4UM�iB��>�HD4a KA,t�8�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L�y&�Yi�HLDH�2�f�@NSi�HLDH�2�v�`NSո�&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,D�%8b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,�18h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��98b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,t�A8h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��M8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A, $�\`d�@np��pnPxdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��pnPq``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��pnPs `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��pnPv@`@p�`�d��D��#�`@.��A8�0<Sa 1A,��U8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a KA,��]8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,t�e8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,4�m8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��u8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��}8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\ah�@np��pn`s `@p�`�d��D��#�`P3��A8�0<Sa A,��\ah�@np��pn`v@`@p�`�d��D��#�`p3��A8�0<Sa .A,��8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8�0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8�0<Sa A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A, $�\`d�@np��n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��n0q``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��n0s `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��n0v@`@p�`�d��D��#�`@.��A8�0<Sa 1A,�� 8�a�� n0�� n0x�@p�d��`�d ��G��<�;#�` B�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a KA,t�8�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,D�%8b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,�18h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��98b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,t�A8h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��M8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A, $�\`d�@np��nPxdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��nPq``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��nPs `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��nPv@`@p�`�d��D��#�`@.��A8�0<Sa 1A,��U8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�Yi�!8MU�F�@NS�iB��>�HD4�iB��>�HD4a KA,��]8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,t�e8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,4�m8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��u8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��}8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\ah�@np��n`s `@p�`�d��D��#�`P3��A8�0<Sa A,��\ah�@np��n`v@`@p�`�d��D��#�`p3��A8�0<Sa .A,��8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�f��4M��4MF��L��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8�0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8�0<Sa A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A, $�\`d�@np��0n0xdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��0n0q``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��0n0s `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��0n0v@`@p�`�d��D��#�`@.��A8�0<Sa 1A,�� 8�a�� n0�� n0x�@p�d��`�d ��G��<�;#�` B�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� �_�O�H#�d� ��4UM�iB��>�HD4a KA,t�8�g�� n0��@n0x�@p�dl@r�d0��`�d@��M��<�A��<�;#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,D�%8b�� n0�� n0q``�dl@q�l�@r�{p�:�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,�18h�� n0��@n0q``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��98b�� n0�� n0s `�dl@q�l�@r�{p�<�A��#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,t�A8h�� n0��@n0s `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��M8b�� n0�� n0v@`�dl@q�l�@r�{p�;�A��#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A, $�\`d�@np��0nPxdl�@q�{@�<�;#�`�p�A8�0<Sd� ��4a A,��\ah�@np��0nPq``@p�`�d��D��#�`�.��A8�0<Sa A,�\ah�@np��0nPs `@p�`�d��D��#�`B�@�0<Sa A,T�\ah�@np��0nPv@`@p�`�d��D��#�`@.��A8�0<Sa 1A,��U8�a�� n0�� nPx�@p�d��`�d ��G��<�;#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�Yi�!8MU�F�@NS�iB��>�HD4�iB��>�HD4a KA,��]8�g�� n0��@nPx�@p�dl@r�d0��`�d@��M��<�A��<�;#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&(��D�D-o� 1!Q˴�i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a .A,t�e8b�� n0�� nPq``�dl@q�l�@r�{p�:�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,4�m8h�� n0��@nPq``�dl@q�d l@s�l�@t�{��:�A��:�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��u8b�� n0�� nPs `�dl@q�l�@r�{p�<�A��#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��}8h�� n0��@nPs `�dl@q�d l@s�l�@t�{��<�A��<�A��#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� nPv@`�dl@q�l�@r�{p�;�A��#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\ah�@np��0n`s `@p�`�d��D��#�`P3��A8�0<Sa A,��\ah�@np��0n`v@`@p�`�d��D��#�`p3��A8�0<Sa .A,��8b�� n0�� n`s `�dl@q�l�@r�{p�9�A��#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a AA,��8h�� n0��@n`s `�dl@q�d l@s�l�@t�{��9�A��9�A��#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa .A,��8b�� n0�� n`v@`�dl@q�l�@r�{p�9�A��#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,$�\jd�@np��Pn0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,��\kh�@np��Pn0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8�0<Sa A,�\kh�@np��Pn0s `@p�`�d��N��>�C9��;�C>�#�`B�@�0<Sa A,T�\kh�@np��Pn0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa 5A,�� 8Wa�� n�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�` B�0�� Bpp0�p�lC �&��L6��L�y&�Yi�!8MU�F�@NS�iB��>�HD4�iB��>�HD4a OA,t�8W(g�� n��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!6��L��&��TV��L�y&�i�HLDH�2�v�`NSո�&��T5o� 1!Q�d� �_�O�H#�d� �_�o�H#�d� �_�_�H#�d� �_��H#�d� ��4�ma 2A,D�%8�b�� n�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,�18�(h�� n��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��98�b�� n�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,t�A8�(h�� n��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��M8�b�� n�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,$�\jd�@np��PnPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,��\kh�@np��PnPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8�0<Sa A,�\kh�@np��PnPs `@p�`�d��N��>�C9��;�C>�#�`B�@�0<Sa A,T�\kh�@np��PnPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa 5A,��U8Wa�� n�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a OA,��]8W(g�� n��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!f�@NSi�HLDH�2�V��L��&(��D�D-�n� ��4U�j�!8MU�&��L�0<Sd� �_�O�H#�d� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a 2A,t�e8�b�� n�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,4�m8�(h�� n��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�f��4MV��L��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��u8�b�� n�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,��}8�(h�� n��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��8�b�� n�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\kh�@np��Pn`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8�0<Sa A,��\kh�@np��Pn`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8�0<Sa 2A,��8�b�� n�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,��8�(h�� n��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��8�b�� n�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,$�\jd�@np��`n0xdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,��\kh�@np��`n0q``@p�`�d��N��>�C9��;�C>�#�`�.��A8�0<Sa A,�\kh�@np��`n0s `@p�`�d��N��>�C9��;�C>�#�`B�@�0<Sa A,T�\kh�@np��`n0v@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa 5A,�� 8Wa�� n0�� n0x�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�` B�0�� Bpp0�p�lC �&��L6��L�y&�Yi�!8MU�F�@NS�iB��>�HD4�iB��>�HD4a OA,t�8W(g�� n0��@n0x�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�`PB�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!��4M6��L��&(��D�D-�n� 1!Q��i�!8MU��0<Sd� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� �_��H#�d� ��4�ma 2A,D�%8�b�� n0�� n0q``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,�18�(h�� n0��@n0q``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��98�b�� n0�� n0s `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�B�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,t�A8�(h�� n0��@n0s `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�B�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��M8�b�� n0�� n0v@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�` C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,$�\jd�@np��`nPxdl�@q�{��<�;��?�C;�C;��<�C?#�`�p�A8�0<Sd� ��4a A,��\kh�@np��`nPq``@p�`�d��N��>�C9��;�C>�#�`�.��A8�0<Sa A,�\kh�@np��`nPs `@p�`�d��N��>�C9��;�C>�#�`B�@�0<Sa A,T�\kh�@np��`nPv@`@p�`�d��N��>�C9��;�C>�#�`@.��A8�0<Sa 5A,��U8Wa�� n0�� nPx�@p�d��`�d ��Q��<�;��?�C;�C;��<�C?#�`@C�0�� Bpp0�p�lC �&��L6��L�y&�i!8Me� ��4UM�iB��>�HD4�iB��>�HD4a OA,��]8W(g�� n0��@nPx�@p�dl@r�d0��`�d@��W��<�A��<�;��?�C;�C;��<�C?#�``C�0�� B0�0� �pppp0۰�lC@� !�6ĐA8!�0<Sd� 1!Q˴[i�HLDH�2�v�`NSո�&(��D�D-�n� ��4U͛hBHLDH�2�jB��>�HD4�i!8Me� �_�o�H#�d� ��4�m[kB��>�HD4�jB��>�HD4a 2A,t�e8�b�� n0�� nPq``�dl@q�l�@r�{�:�A��>�C9��;�C>�#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,4�m8�(h�� n0��@nPq``�dl@q�d l@s�l�@t�{p�:�A��:�A��>�C9��;�C>�#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��u8�b�� n0�� nPs `�dl@q�l�@r�{��<�A��>�C9��;�C>�#�`�C�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,��}8�(h�� n0��@nPs `�dl@q�d l@s�l�@t�{p��<�A��<�A��>�C9��;�C>�#�`�C�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�Yi�HLDH�2�F��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��8�b�� n0�� nPv@`�dl@q�l�@r�{�;�A��>�C9��;�C>�#�`F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,d�\kh�@np��`n`s `@p�`�d��N��>�C9��;�C>�#�`P3��A8�0<Sa A,��\kh�@np��`n`v@`@p�`�d��N��>�C9��;�C>�#�`p3��A8�0<Sa 2A,��8�b�� n0�� n`s `�dl@q�l�@r�{��9�A��>�C9��;�C>�#�`0F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a EA,��8�(h�� n0��@n`s `�dl@q�d l@s�l�@t�{p��9�A��9�A��>�C9��;�C>�#�``F�0�� B0�0� �0�p�lC@� !�6ĐA86��L�y&�i�HLDH�2�V��L��&4~!>�#�DD��&4~!��#�DD��&$��D�D-��&4~!~�#�DD��&4~!��#�DDa 2A,��8�b�� n0�� n`v@`�dl@q�l�@r�{�9�A��>�C9��;�C>�#�`�F�0�� B0�P�lC �&��L6��L�y&�iB��>�HD4YiB��>�HD4a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,��\id�@np�� n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np�� n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np�� n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np�� n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��=��؃��<��A/��<��P�� 0�0' ' #�`�F�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,�\}��<Ѓ�;��8�;��=��<��A/��L��P�� 0�0� �0�PG G G G # �`�F�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,$�\t��<Ѓ�;��8�;��=��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,D�\~��<Ѓ�;��8�;��=��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`0G�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,d�\t��<Ѓ�;��8�;��=��؃��<́��.��>�A��A?��l l l�� 0�0#�`PGC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��=��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`pG�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��<��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,��\id�@np�� nPx�@p�ld{а��>�C9��;�C>�'#�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np�� nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np�� nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np�� nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��=��؃�A=��A/��<��P�� 0�0''#�`�G�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,��\}��<Ѓ�;��8�;��=��A=��A/��L��P�� 0�0� �0�PGGGG# �`�G�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,�\t��<Ѓ�;��8�;��=��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,$�\~��<Ѓ�;��8�;��=��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,D�\t��<Ѓ�;��8�;��=��؃�A=́��.��>�A��A?��l l l�� 0�0#�`0JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,d�\~��<Ѓ�;��8�;��=��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`PJ�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`pJC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,t�\jh�@np�� n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8�0<Sd� E,a A,��\jh�@np�� n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8�0<Sd� E,a ,A,��\t��<Ѓ�;��8�;��=��؃��9́��.��>�A��A?��l`l`l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��=��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,��\id�@np��pn0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��pn0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��pn0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��pn0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��8��؃��<��A/��<��P�� 0�0' ' #�`�F�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,�\}��<Ѓ�;��8�;��8��<��A/��L��P�� 0�0� �0�PG G G G # �`�F�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,$�\t��<Ѓ�;��8�;��8��؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,D�\~��<Ѓ�;��8�;��8��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`0G�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,d�\t��<Ѓ�;��8�;��8��؃��<́��.��>�A��A?��l l l�� 0�0#�`PGC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��8��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`pG�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��<��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,��\id�@np��pnPx�@p�ld{а��>�C9��;�C>�'#�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��pnPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��pnPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��pnPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��8��؃�A=��A/��<��P�� 0�0''#�`�G�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,��\}��<Ѓ�;��8�;��8��A=��A/��L��P�� 0�0� �0�PGGGG# �`�G�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,�\t��<Ѓ�;��8�;��8��؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,$�\~��<Ѓ�;��8�;��8��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,D�\t��<Ѓ�;��8�;��8��؃�A=́��.��>�A��A?��l l l�� 0�0#�`0JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,d�\~��<Ѓ�;��8�;��8��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`PJ�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃�A=��.��>�A��A?��l�l�l�� 0�0#�`pJC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,t�\jh�@np��pn`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8�0<Sd� E,a A,��\jh�@np��pn`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8�0<Sd� E,a ,A,��\t��<Ѓ�;��8�;��8��؃��9́��.��>�A��A?��l`l`l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��8��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8��؃��9��.��>�A��A?��l@l@l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,��\id�@np��0n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��0n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��0n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��0n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��8̃؃��<��A/��<��P�� 0�0' ' #�`�F�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,�\}��<Ѓ�;��8�;��8̃��<��A/��L��P�� 0�0� �0�PG G G G # �`�F�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,$�\t��<Ѓ�;��8�;��8̃؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,D�\~��<Ѓ�;��8�;��8̃��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`0G�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,d�\t��<Ѓ�;��8�;��8̃؃��<́��.��>�A��A?��l l l�� 0�0#�`PGC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��8̃��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`pG�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��<��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,��\id�@np��0nPx�@p�ld{а��>�C9��;�C>�'#�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��0nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��0nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��0nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��8̃؃�A=��A/��<��P�� 0�0''#�`�G�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,��\}��<Ѓ�;��8�;��8̃��A=��A/��L��P�� 0�0� �0�PGGGG# �`�G�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,�\t��<Ѓ�;��8�;��8̃؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,$�\~��<Ѓ�;��8�;��8̃��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,D�\t��<Ѓ�;��8�;��8̃؃�A=́��.��>�A��A?��l l l�� 0�0#�`0JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,d�\~��<Ѓ�;��8�;��8̃��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`PJ�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃�A=��.��>�A��A?��l�l�l�� 0�0#�`pJC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,t�\jh�@np��0n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8�0<Sd� E,a A,��\jh�@np��0n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8�0<Sd� E,a ,A,��\t��<Ѓ�;��8�;��8̃؃��9́��.��>�A��A?��l`l`l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��8̃��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��8̃؃��9��.��>�A��A?��l@l@l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,��\id�@np��@n0x�@p�ld{а��>�C9��;�C>�' #�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��@n0q``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��@n0s `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��@n0v@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃��<��A/��<��P�� 0�0' ' #�`�F�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,�\}��<Ѓ�;��8�;��=Ѓ��<��A/��L��P�� 0�0� �0�PG G G G # �`�F�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,$�\t��<Ѓ�;��8�;��=Ѓ؃��<ā��.��>�A��A?��l�l�l�� 0�0#�`GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,D�\~��<Ѓ�;��8�;��=Ѓ��<ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`0G�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,d�\t��<Ѓ�;��8�;��=Ѓ؃��<́��.��>�A��A?��l l l�� 0�0#�`PGC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��<́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`pG�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��<��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,��\id�@np��@nPx�@p�ld{а��>�C9��;�C>�'#�`�F`8�0<Sd� ��4�}&�Pa A,t�\jh�@np��@nPq``�d��`@q� ;�:��?�C;�C;��<�C?#�``6� `8�0<Sd� E,a A,��\jh�@np��@nPs `�d��`@q� ;��<��?�C;�C;��<�C?#�`�6� `8�0<Sd� E,a A,4�\jh�@np��@nPv@`�d��`@q� ;�;��?�C;�C;��<�C?#�` 6� `8�0<Sd� E,a 0A,��\s��<Ѓ�;��8�;��=Ѓ؃�A=��A/��<��P�� 0�0''#�`�G�!&�0U�B~!~��d� ��4n� ��4�y&�Yh�TE��O\S�}&�Pa DA,��\}��<Ѓ�;��8�;��=Ѓ��A=��A/��L��P�� 0�0� �0�PGGGG# �`�G�BB��@6�0U�B~!��d� S,��>qM Lf�PNS�v�PNS�V�@NSj�!8MU��0<Sd� E,a� S,��>qM L&�0U�B~!~��a ,A,�\t��<Ѓ�;��8�;��=Ѓ؃�A=ā��.��>�A��A?��l�l�l�� 0�0#�`�GC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,$�\~��<Ѓ�;��8�;��=Ѓ��A=ā��.��>�A��A?��l�l�l�l�l�� 0�0� �0�P# �`J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,D�\t��<Ѓ�;��8�;��=Ѓ؃�A=́��.��>�A��A?��l l l�� 0�0#�`0JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,d�\~��<Ѓ�;��8�;��=Ѓ��A=́��.��>�A��A?��l l l l l�� 0�0� �0�P# �`PJ�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃�A=��.��>�A��A?��l�l�l�� 0�0#�`pJC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,t�\jh�@np��@n`s `�d��`@q� ;��9��?�C;�C;��<�C?#�``:� `8�0<Sd� E,a A,��\jh�@np��@n`v@`�d��`@q� ;�9��?�C;�C;��<�C?#�`�:� `8�0<Sd� E,a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9́��.��>�A��A?��l`l`l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a ;A,��\~��<Ѓ�;��8�;��=Ѓ��9́��.��>�A��A?��l`l`l`l`l�� 0�0� �0�P# �`�J�AB��@F�0U�B~!��d� �3E��PT�B�0U�B~!>��d� S,��>qM L6�0U�B~!��a ,A,��\t��<Ѓ�;��8�;��=Ѓ؃��9��.��>�A��A?��l@l@l�� 0�0#�`�JC!�0<Sd� S,��>qM L&�0U�B~!~��d� E,a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�� y&�c� �t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�� y&�c� �t �a A,��\ix�� n s `@p�`@q�`@r�`@s�H��#�`�1�� y&�c� �t �a A,��\j|��0rdldld ld0{��<��<��<��<#�`�1�� y&�c� �t �a �� ܇HA�@��PT�B�@0��a �� ܇HA�@�@0��gBQEa �� ?A�@�@0��gBQEa �� ?A�@�@0��gBQEa �� ?A�@�@0��gBQEa �� ?A�@��PT�B�@0��a ��A� ��>#�`p#1��@0��h!�b� E,a ��A� � >#�`; 1��@0��h!�b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a �� =�hB�&�İQ�C8�@<1�g��b� E,a ��A� ��>#�`�;�@� 0� ��@�@�@<1�g��b� �,�}&�Pa ��A� �I>#�`�;�@� 0� ��@�@�@<1�g��b� �,�}&�Pa A, $�\U�C8��;�;�C9��A #�`+��@� 4a �� ;B�A8�0T�b� 0!I��p�a� �,a �� G:A�@�@0��h!�b� E,a �� 9A�@�@0��h!�b� E,a �� G9A�@�@0��h!�b� E,a �� 8A�@�@0��h!�b� E,a �� G8A�@�@0��h!�b� E,a �� /A�@�@0��h!�b� E,a �� G/A�@�@0��h!�b� E,a �� .A�@�@0��h!�b� E,a �� G.A�@�@0��h!�b� E,a ��A� � ��0<Sa ��A� �J-��0<Sa �� -A�@�0<Sd� �D1��a ��A� �A,��0<Sa A,��7�� D�6F�A8�� %MD��3��i��>�HD4��&<~!~�#�DD��g��L��&��Ta ��A� �O+��Da A,@� 0#�` '�0��B��1�@�lC0� � d F��4MV��4Mf��4Ma A,@1� 0#�` '�0��0bP ru0�@�lC0� � d F��4MV��4Mf��4Ma A,@IEY�3#�` '�@��1bP r�l1� ��6��A8 V��4Mf��4Mv��4Ma A,@a�Fq�3#�` '�@��1bP r�l1� ��6��A8 V��4Mf��4Mv��4Ma A,@ #�` '�da ��A� ��*�Aa ��A� � )�a ��A� � )�a A,@ #�`+� a ��A� � (�a �� (A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��a �� A�@��0T�b� 0!I�@��q �2"�^�5��h�`�6�j�b�68�m�d�6X�y�f�7x�|�h�7��~�j�8�扃l�8ؖ��n�8��p�9��r�98睃t�9X��v�:x��x�:��z�;�繃|�;ؗ��~�;��ȃ��<�˃��<8�΃��<X�؃��=x�ۃ��=��߃��>��ꃌ�>��탎�>�`��?��?8` ��@X��Ax ��A��)��B� /��C��<��C��J��D�O��E8@]��FX�j��Fx x��G��}��H��I�`��J�P��L�߄��OH ��O��O� ��O��OH ��O��P� ��P��P( ��Ph��P�0��Q��Q(P��Qh��Q��Q�(��R(�)��Rh+��R��,��R� .��R((�/��Sh(09��S�(�:��S�(@<��S()�=��Sh)@?��T�)�H��T�)PJ��T(*�K��Th*pM��T�*0O��T�*�X��U(+0Z��Uh+�[��U�+0]��U�+�^��U(,�h��Vh,j��V�,�k��V�, m��V(-�n��Vh-Px��W�-�y��W�-`{��W(.�|��Wh.`~��W�.��X�.��X/`��XH/p��Yx/��Y�/��Y�/��Z�/@��Z8��Z88��ZX8`��[�8຅��[�8`��\�8Ʌ��\9�ʅ��\89@̅��\X9 ΅��\�9؅��]�9 څ��]�9܅��]8:�݅��]x:�߅��^�:�酭�^�:�녱�^8;`텵�^x;@�_�;@��_�;@��_8<@��_x<��`=��a�=��bX>�-��c�>�<��d�?�K��d�?@M��d�?�N��d(H�X��ehH�Z��e�H�\��e�Hp^��e(IPh��fhI0j��f�Il��f�I�m��f(J�o��ghJ�y��g�J�{��g�J�}��g�J��h(K��hhK��h�Kp��h�KP��i(L0��ihL��i�L��i�LО��i(M��jhM��j�Mp��j�MP��j(N0��khN0��k�NP��k�Np��k(OpȆ��lhOPʆ��l�O0̆��l�OΆ��l(X�φ��mhX�ن��m�X�ۆ��m�X�݆��m(Yp߆��nhYP醘�n�Y0놜�n�Y토�n(Z��nhZ ��o�Z0��o�Z��o([��oH[��oh[0��p�[p ��p�[� ��p�[� ��p\��qH\��qx\��q�\@��q�\ (��r�\�+��r]/��s(]�:��sh]�>��t�]0K��t�]@Z��u�]�k��w�]}��x^��y(^��zH^��{h^�ˇ��}�^�݇��~�^`��^p ��^@��_��ׁ(_ ��H_p*��ڂh_p:��׃�_�I��ʄ�_�X��_�]��_�h��h�m��(h�}��Hh��hh��ډ�hP��h��h0��΋�h��i��(i`��Hi`��hi��i��i��i��ؐ�i��ؑj�(��˒(j�/��Hj�>��hj@J��Ԕ�jpY��Ǖ�j ^��j�i��jm��kPx��(kP��Hk��hk��kp��Қ�k��k��kp��l��(l��Hl0��hl`��l ��l��l��l@��m@+��(mp:��ףHm�I��ʤhm`N��m�Y��m�^��m`n��m�}��n��(n ��HnP��hn0��n��n��ܫ�n��n ��o��(o��Ho��hop��o� ��o��o�-��op=��x0I��Ǵ(x�X��µHx^��hx�i��x�m��xpy��η�x��Ӹ�x��ιyP��ɺ(y��Hy@��»hy��ʼ�y��Ž�y��y@��y��z� ��(z��Hz��hz(��z,��z�;��z@J��z�X��{�]��({�i��H{o��h{@~��{��{��{��{P��|0��(|@��H|P��h|P��|��|��|0��| ��} ��(}@��H}�)��h}08��}�>��}�J��}�Y��}p^��~�j��(~�n��H~�z��h~��~��~p��~`��~`��(��Hp��h �� (��H�P-��h�<��J��O��Ȉ�[��h��x��(��~��H�`��h�`��ȉ ��(��H�@��x��0��p��(�0��H��ȋ��P��H��0��(� ��X� ��x� ��ȍ�� 8��x��]}��KthreadIdxblockIdxblockDimgridDimwarpSize_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9all_hostsE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_35_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_37_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_50_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_52_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_53_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_60_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_61_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_62_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_70_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_72_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_75_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_80_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_86_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_87_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_89_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail9sm_90_bitE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target6detail11all_devicesE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target7is_hostE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target9is_deviceE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target10any_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target9no_targetE_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_35E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_37E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_50E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_52E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_53E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_60E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_61E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_62E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_70E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_72E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_75E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_80E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_86E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_87E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_89E_ZN56_INTERNAL_00000000_25_nvrtc_builtins_compute_90_0fd890352nv6target5sm_90E__cudaCDP2DeviceGetAttribute__cudaCDP2DeviceGetLimit__cudaCDP2DeviceGetCacheConfig__cudaCDP2DeviceGetSharedMemConfig__cudaCDP2GetLastError__cudaCDP2PeekAtLastError__cudaCDP2GetErrorString__cudaCDP2GetErrorName__cudaCDP2GetDeviceCount__cudaCDP2GetDevice__cudaCDP2StreamCreateWithFlags__cudaCDP2StreamDestroy__cudaCDP2StreamWaitEvent__cudaCDP2StreamWaitEvent_ptsz__cudaCDP2EventCreateWithFlags__cudaCDP2EventRecord__cudaCDP2EventRecord_ptsz__cudaCDP2EventRecordWithFlags__cudaCDP2EventRecordWithFlags_ptsz__cudaCDP2EventDestroy__cudaCDP2FuncGetAttributes__cudaCDP2Free__cudaCDP2Malloc__cudaCDP2MemcpyAsync__cudaCDP2MemcpyAsync_ptsz__cudaCDP2Memcpy2DAsync__cudaCDP2Memcpy2DAsync_ptsz__cudaCDP2Memcpy3DAsync__cudaCDP2Memcpy3DAsync_ptsz__cudaCDP2MemsetAsync__cudaCDP2MemsetAsync_ptsz__cudaCDP2Memset2DAsync__cudaCDP2Memset2DAsync_ptsz__cudaCDP2Memset3DAsync__cudaCDP2Memset3DAsync_ptsz__cudaCDP2RuntimeGetVersion__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessor__cudaCDP2OccupancyMaxActiveBlocksPerMultiprocessorWithFlags__cudaCDP2GetParameterBuffer__cudaCDP2GetParameterBufferV2__cudaCDP2LaunchDevice_ptsz__cudaCDP2LaunchDeviceV2_ptsz__cudaCDP2LaunchDevice__cudaCDP2LaunchDeviceV2mallocfreellvm.memcpy.p0i8.p0i8.i64llvm.memset.p0i8.i64llvm.nvvm.associate.access.property.p0i8.p0i8llvm.nvvm.memcpy.async.p3i8.p1i8__nv_aligned_device_malloc_impl_Z10make_char1a_Z11make_uchar1h_Z10make_char2aa_Z11make_uchar2hh_Z10make_char3aaa_Z11make_uchar3hhh_Z10make_char4aaaa_Z11make_uchar4hhhh_Z11make_short1s_Z12make_ushort1t_Z11make_short2ss_Z12make_ushort2tt_Z11make_short3sss_Z12make_ushort3ttt_Z11make_short4ssss_Z12make_ushort4tttt_Z9make_int1i_Z10make_uint1j_Z9make_int2ii_Z10make_uint2jj_Z9make_int3iii_Z10make_uint3jjj_Z9make_int4iiii_Z10make_uint4jjjj_Z10make_long1l_Z11make_ulong1m_Z10make_long2ll_Z11make_ulong2mm_Z10make_long3lll_Z11make_ulong3mmm_Z10make_long4llll_Z11make_ulong4mmmm_Z11make_float1f_Z11make_float2ff_Z11make_float3fff_Z11make_float4ffff_Z14make_longlong1x_Z15make_ulonglong1y_Z14make_longlong2xx_Z15make_ulonglong2yy_Z14make_longlong3xxx_Z15make_ulonglong3yyy_Z14make_longlong4xxxx_Z15make_ulonglong4yyyy_Z12make_double1d_Z12make_double2dd_Z12make_double3ddd_Z12make_double4dddd__vimax_s32_relu__vimax_s16x2_relu__vimin_s32_relu__vimin_s16x2_relu__vimax3_s32__vimax3_s16x2__vimax3_u32__vimax3_u16x2__vimin3_s32__vimin3_s16x2__vimin3_u32__vimin3_u16x2__vimax3_s32_relu__vimax3_s16x2_relu__vimin3_s32_relu__vimin3_s16x2_relu__viaddmax_s32__viaddmax_s16x2__viaddmax_u32__viaddmax_u16x2__viaddmin_s32__viaddmin_s16x2__viaddmin_u32__viaddmin_u16x2__viaddmax_s32_relu__viaddmax_s16x2_relu__viaddmin_s32_relu__viaddmin_s16x2_relu__vibmax_s32__vibmax_u32__vibmin_s32__vibmin_u32__vibmax_s16x2__vibmax_u16x2__vibmin_s16x2__vibmin_u16x2_Z7signbitf__signbitf_Z7signbitd__signbit_Z7signbite__signbitl_Z8isfinitef__finitef_Z8isfinited__finite_Z8isfinitee__finitel_Z5isnanf__isnanf_Z5isnand__isnan_Z5isnane__isnanl_Z5isinff__isinff_Z5isinfd__isinf_Z5isinfe__isinfl_Z3absxllabs_Z3absl_Z3absffabsf_Z3absdfabs_Z4fabsf_Z4ceilfceilf_Z5floorffloorf_Z4sqrtfsqrtf_Z3powffpowf_Z3powfi_Z5powiffi_Z3powdi_Z4powidi_Z3logflogf_Z5log10flog10f_Z4fmodfffmodf_Z4modffPfmodff_Z3expfexpf_Z5frexpfPifrexpf_Z5ldexpfildexpf_Z4asinfasinf_Z3sinfsinf_Z4sinhfsinhf_Z4acosfacosf_Z3cosfcosf_Z4coshfcoshf_Z4atanfatanf_Z5atan2ffatan2f_Z3tanftanf_Z4tanhftanhf_Z4logbflogbf_Z5ilogbfilogbf_Z6scalbnfiscalbnf_Z7scalblnflscalblnf_Z4exp2fexp2f_Z5expm1fexpm1f_Z4log2flog2f_Z5log1pflog1pf_Z5acoshfacoshf_Z5asinhfasinhf_Z5atanhfatanhf_Z5hypotffhypotf_Z4cbrtfcbrtf_Z3erfferff_Z4erfcferfcf_Z6lgammaflgammaf_Z6tgammaftgammaf_Z8copysignffcopysignf_Z9nextafterffnextafterf_Z9remainderffremainderf_Z6remquoffPiremquof_Z5roundfroundf_Z6lroundflroundf_Z7llroundfllroundf_Z5truncftruncf_Z4rintfrintf_Z5lrintflrintf_Z6llrintfllrintf_Z9nearbyintfnearbyintf_Z4fdimfffdimf_Z3fmaffffmaf_Z4fmaxfffmaxf_Z4fminfffminf_Z5exp10fexp10f_Z5rsqrtfrsqrtf_Z5rcbrtfrcbrtf_Z5sinpifsinpif_Z5cospifcospif_Z8sincospifPfS_sincospif_Z6sincosfPfS_sincosf_Z2j0fj0f_Z2j1fj1f_Z2jnifjnf_Z2y0fy0f_Z2y1fy1f_Z2ynifynf_Z13cyl_bessel_i0fcyl_bessel_i0f_Z13cyl_bessel_i1fcyl_bessel_i1f_Z6erfinvferfinvf_Z7erfcinvferfcinvf_Z10normcdfinvfnormcdfinvf_Z7normcdffnormcdff_Z5erfcxferfcxf_Z8copysigndfcopysign_Z8copysignfd_Z3minjjumin_Z3minij_Z3minji_Z3minllminllmin_Z3minmmullmin_Z3minlm_Z3minml_Z3minxx_Z3minyy_Z3minxy_Z3minyx_Z3minff_Z3minddfmin_Z3minfd_Z3mindf_Z3maxjjumax_Z3maxij_Z3maxji_Z3maxllmaxllmax_Z3maxmmullmax_Z3maxlm_Z3maxml_Z3maxxx_Z3maxyy_Z3maxxy_Z3maxyx_Z3maxff_Z3maxddfmax_Z3maxfd_Z3maxdf_Z9atomicAddPii__iAtomicAdd_Z9atomicAddPjj__uAtomicAdd_Z9atomicSubPii_Z9atomicSubPjj_Z10atomicExchPii__iAtomicExch_Z10atomicExchPjj__uAtomicExch_Z10atomicExchPff__fAtomicExch_Z9atomicMinPii__iAtomicMin_Z9atomicMinPjj__uAtomicMin_Z9atomicMaxPii__iAtomicMax_Z9atomicMaxPjj__uAtomicMax_Z9atomicIncPjj__uAtomicInc_Z9atomicDecPjj__uAtomicDec_Z9atomicAndPii__iAtomicAnd_Z9atomicAndPjj__uAtomicAnd_Z8atomicOrPii__iAtomicOr_Z8atomicOrPjj__uAtomicOr_Z9atomicXorPii__iAtomicXor_Z9atomicXorPjj__uAtomicXor_Z9atomicCASPiii__iAtomicCAS_Z9atomicCASPjjj__uAtomicCAS_Z9atomicAddPyy__ullAtomicAdd_Z10atomicExchPyy__ullAtomicExch_Z9atomicCASPyyy__ullAtomicCAS_Z3anyb__any_Z3allb__all_Z3fmaddd13cudaRoundMode__fma_rz__fma_ru__fma_rd__fma_rn_Z4dmuldd13cudaRoundMode__dmul_rz__dmul_ru__dmul_rd__dmul_rn_Z4dadddd13cudaRoundMode__dadd_rz__dadd_ru__dadd_rd__dadd_rn_Z4dsubdd13cudaRoundMode__dsub_rz__dsub_ru__dsub_rd__dsub_rn_Z10double2intd13cudaRoundMode__double2int_rn__double2int_ru__double2int_rd__double2int_rz_Z11double2uintd13cudaRoundMode__double2uint_rn__double2uint_ru__double2uint_rd__double2uint_rz_Z9double2lld13cudaRoundMode__double2ll_rn__double2ll_ru__double2ll_rd__double2ll_rz_Z10double2ulld13cudaRoundMode__double2ull_rn__double2ull_ru__double2ull_rd__double2ull_rz_Z9ll2doublex13cudaRoundMode__ll2double_rz__ll2double_ru__ll2double_rd__ll2double_rn_Z10ull2doubley13cudaRoundMode__ull2double_rz__ull2double_ru__ull2double_rd__ull2double_rn_Z10int2doublei13cudaRoundMode_Z11uint2doublej13cudaRoundMode_Z12float2doublef13cudaRoundMode_Z9atomicAddPff__fAtomicAdd_Z9atomicMinPxx__illAtomicMin_Z9atomicMaxPxx__illAtomicMax_Z9atomicAndPxx__llAtomicAnd_Z8atomicOrPxx__llAtomicOr_Z9atomicXorPxx__llAtomicXor_Z9atomicMinPyy__ullAtomicMin_Z9atomicMaxPyy__ullAtomicMax_Z9atomicAndPyy__ullAtomicAnd_Z8atomicOrPyy__ullAtomicOr_Z9atomicXorPyy__ullAtomicXor_Z9atomicAddPdd__dAtomicAdd_Z15atomicAdd_blockPii__iAtomicAdd_block_Z16atomicAdd_systemPii__iAtomicAdd_system_Z15atomicAdd_blockPjj__uAtomicAdd_block_Z16atomicAdd_systemPjj__uAtomicAdd_system_Z15atomicAdd_blockPyy__ullAtomicAdd_block_Z16atomicAdd_systemPyy__ullAtomicAdd_system_Z15atomicAdd_blockPff__fAtomicAdd_block_Z16atomicAdd_systemPff__fAtomicAdd_system_Z15atomicAdd_blockPdd__dAtomicAdd_block_Z16atomicAdd_systemPdd__dAtomicAdd_system_Z15atomicSub_blockPii_Z16atomicSub_systemPii_Z15atomicSub_blockPjj_Z16atomicSub_systemPjj_Z16atomicExch_blockPii__iAtomicExch_block_Z17atomicExch_systemPii__iAtomicExch_system_Z16atomicExch_blockPjj__uAtomicExch_block_Z17atomicExch_systemPjj__uAtomicExch_system_Z16atomicExch_blockPyy__ullAtomicExch_block_Z17atomicExch_systemPyy__ullAtomicExch_system_Z16atomicExch_blockPff__fAtomicExch_block_Z17atomicExch_systemPff__fAtomicExch_system_Z15atomicMin_blockPii__iAtomicMin_block_Z16atomicMin_systemPii__iAtomicMin_system_Z15atomicMin_blockPxx__illAtomicMin_block_Z16atomicMin_systemPxx__illAtomicMin_system_Z15atomicMin_blockPjj__uAtomicMin_block_Z16atomicMin_systemPjj__uAtomicMin_system_Z15atomicMin_blockPyy__ullAtomicMin_block_Z16atomicMin_systemPyy__ullAtomicMin_system_Z15atomicMax_blockPii__iAtomicMax_block_Z16atomicMax_systemPii__iAtomicMax_system_Z15atomicMax_blockPxx__illAtomicMax_block_Z16atomicMax_systemPxx__illAtomicMax_system_Z15atomicMax_blockPjj__uAtomicMax_block_Z16atomicMax_systemPjj__uAtomicMax_system_Z15atomicMax_blockPyy__ullAtomicMax_block_Z16atomicMax_systemPyy__ullAtomicMax_system_Z15atomicInc_blockPjj__uAtomicInc_block_Z16atomicInc_systemPjj__uAtomicInc_system_Z15atomicDec_blockPjj__uAtomicDec_block_Z16atomicDec_systemPjj__uAtomicDec_system_Z15atomicCAS_blockPiii__iAtomicCAS_block_Z16atomicCAS_systemPiii__iAtomicCAS_system_Z15atomicCAS_blockPjjj__uAtomicCAS_block_Z16atomicCAS_systemPjjj__uAtomicCAS_system_Z15atomicCAS_blockPyyy__ullAtomicCAS_block_Z16atomicCAS_systemPyyy__ullAtomicCAS_system_Z15atomicAnd_blockPii__iAtomicAnd_block_Z16atomicAnd_systemPii__iAtomicAnd_system_Z15atomicAnd_blockPxx__llAtomicAnd_block_Z16atomicAnd_systemPxx__llAtomicAnd_system_Z15atomicAnd_blockPjj__uAtomicAnd_block_Z16atomicAnd_systemPjj__uAtomicAnd_system_Z15atomicAnd_blockPyy__ullAtomicAnd_block_Z16atomicAnd_systemPyy__ullAtomicAnd_system_Z14atomicOr_blockPii__iAtomicOr_block_Z15atomicOr_systemPii__iAtomicOr_system_Z14atomicOr_blockPxx__llAtomicOr_block_Z15atomicOr_systemPxx__llAtomicOr_system_Z14atomicOr_blockPjj__uAtomicOr_block_Z15atomicOr_systemPjj__uAtomicOr_system_Z14atomicOr_blockPyy__ullAtomicOr_block_Z15atomicOr_systemPyy__ullAtomicOr_system_Z15atomicXor_blockPii__iAtomicXor_block_Z16atomicXor_systemPii__iAtomicXor_system_Z15atomicXor_blockPxx__llAtomicXor_block_Z16atomicXor_systemPxx__llAtomicXor_system_Z15atomicXor_blockPjj__uAtomicXor_block_Z16atomicXor_systemPjj__uAtomicXor_system_Z15atomicXor_blockPyy__ullAtomicXor_block_Z16atomicXor_systemPyy__ullAtomicXor_system_Z6ballotb__ballot_Z17syncthreads_countb__syncthreads_count_Z15syncthreads_andb__syncthreads_and_Z14syncthreads_orb__syncthreads_or_Z10__isGlobalPKvllvm.nvvm.isspacep.global_Z10__isSharedPKvllvm.nvvm.isspacep.shared_Z12__isConstantPKvllvm.nvvm.isspacep.const_Z9__isLocalPKvllvm.nvvm.isspacep.local_Z16__isGridConstantPKvllvm.nvvm.isspacep.grid_const_Z24__cvta_generic_to_globalPKv_Z24__cvta_generic_to_sharedPKv_Z26__cvta_generic_to_constantPKv_Z23__cvta_generic_to_localPKv_Z24__cvta_global_to_genericm_Z24__cvta_shared_to_genericm_Z26__cvta_constant_to_genericm_Z23__cvta_local_to_genericm_Z31__cvta_generic_to_grid_constantPKv_Z31__cvta_grid_constant_to_genericm_Z5__fnsjjillvm.nvvm.fns_Z14__barrier_syncjllvm.nvvm.barrier.sync_Z20__barrier_sync_countjjllvm.nvvm.barrier.sync.cnt_Z10__syncwarpjllvm.nvvm.bar.warp.sync_Z10__all_syncjillvm.nvvm.vote.sync_Z10__any_syncji_Z10__uni_syncji_Z13__ballot_syncji_Z12__activemaskv_Z11__shfl_syncjiiillvm.nvvm.shfl.sync.i32_Z11__shfl_syncjjii_Z14__shfl_up_syncjiji_Z14__shfl_up_syncjjji_Z16__shfl_down_syncjiji_Z16__shfl_down_syncjjji_Z15__shfl_xor_syncjiii_Z15__shfl_xor_syncjjii_Z11__shfl_syncjfii__float_as_int__int_as_float_Z14__shfl_up_syncjfji_Z16__shfl_down_syncjfji_Z15__shfl_xor_syncjfii_Z11__shfl_syncjxii_Z11__shfl_syncjyii_Z14__shfl_up_syncjxji_Z14__shfl_up_syncjyji_Z16__shfl_down_syncjxji_Z16__shfl_down_syncjyji_Z15__shfl_xor_syncjxii_Z15__shfl_xor_syncjyii_Z11__shfl_syncjdii_Z14__shfl_up_syncjdji_Z16__shfl_down_syncjdji_Z15__shfl_xor_syncjdii_Z11__shfl_syncjlii_Z11__shfl_syncjmii_Z14__shfl_up_syncjlji_Z14__shfl_up_syncjmji_Z16__shfl_down_syncjlji_Z16__shfl_down_syncjmji_Z15__shfl_xor_syncjlii_Z15__shfl_xor_syncjmii_Z5__ldgPKl_Z5__ldgPKm_Z5__ldgPKc_Z5__ldgPKa_Z5__ldgPKs_Z5__ldgPKi_Z5__ldgPKx_Z5__ldgPK5char2_Z5__ldgPK5char4_Z5__ldgPK6short2_Z5__ldgPK6short4_Z5__ldgPK4int2_Z5__ldgPK4int4_Z5__ldgPK9longlong2_Z5__ldgPKh_Z5__ldgPKt_Z5__ldgPKj_Z5__ldgPKy_Z5__ldgPK6uchar2_Z5__ldgPK6uchar4_Z5__ldgPK7ushort2_Z5__ldgPK7ushort4_Z5__ldgPK5uint2_Z5__ldgPK5uint4_Z5__ldgPK10ulonglong2_Z5__ldgPKf_Z5__ldgPKd_Z5__ldgPK6float2_Z5__ldgPK6float4_Z5__ldgPK7double2_Z6__ldcgPKl_Z6__ldcgPKm_Z6__ldcgPKc_Z6__ldcgPKa_Z6__ldcgPKs_Z6__ldcgPKi_Z6__ldcgPKx_Z6__ldcgPK5char2_Z6__ldcgPK5char4_Z6__ldcgPK6short2_Z6__ldcgPK6short4_Z6__ldcgPK4int2_Z6__ldcgPK4int4_Z6__ldcgPK9longlong2_Z6__ldcgPKh_Z6__ldcgPKt_Z6__ldcgPKj_Z6__ldcgPKy_Z6__ldcgPK6uchar2_Z6__ldcgPK6uchar4_Z6__ldcgPK7ushort2_Z6__ldcgPK7ushort4_Z6__ldcgPK5uint2_Z6__ldcgPK5uint4_Z6__ldcgPK10ulonglong2_Z6__ldcgPKf_Z6__ldcgPKd_Z6__ldcgPK6float2_Z6__ldcgPK6float4_Z6__ldcgPK7double2_Z6__ldcaPKl_Z6__ldcaPKm_Z6__ldcaPKc_Z6__ldcaPKa_Z6__ldcaPKs_Z6__ldcaPKi_Z6__ldcaPKx_Z6__ldcaPK5char2_Z6__ldcaPK5char4_Z6__ldcaPK6short2_Z6__ldcaPK6short4_Z6__ldcaPK4int2_Z6__ldcaPK4int4_Z6__ldcaPK9longlong2_Z6__ldcaPKh_Z6__ldcaPKt_Z6__ldcaPKj_Z6__ldcaPKy_Z6__ldcaPK6uchar2_Z6__ldcaPK6uchar4_Z6__ldcaPK7ushort2_Z6__ldcaPK7ushort4_Z6__ldcaPK5uint2_Z6__ldcaPK5uint4_Z6__ldcaPK10ulonglong2_Z6__ldcaPKf_Z6__ldcaPKd_Z6__ldcaPK6float2_Z6__ldcaPK6float4_Z6__ldcaPK7double2_Z6__ldcsPKl_Z6__ldcsPKm_Z6__ldcsPKc_Z6__ldcsPKa_Z6__ldcsPKs_Z6__ldcsPKi_Z6__ldcsPKx_Z6__ldcsPK5char2_Z6__ldcsPK5char4_Z6__ldcsPK6short2_Z6__ldcsPK6short4_Z6__ldcsPK4int2_Z6__ldcsPK4int4_Z6__ldcsPK9longlong2_Z6__ldcsPKh_Z6__ldcsPKt_Z6__ldcsPKj_Z6__ldcsPKy_Z6__ldcsPK6uchar2_Z6__ldcsPK6uchar4_Z6__ldcsPK7ushort2_Z6__ldcsPK7ushort4_Z6__ldcsPK5uint2_Z6__ldcsPK5uint4_Z6__ldcsPK10ulonglong2_Z6__ldcsPKf_Z6__ldcsPKd_Z6__ldcsPK6float2_Z6__ldcsPK6float4_Z6__ldcsPK7double2_Z6__ldluPKl_Z6__ldluPKm_Z6__ldluPKc_Z6__ldluPKa_Z6__ldluPKs_Z6__ldluPKi_Z6__ldluPKx_Z6__ldluPK5char2_Z6__ldluPK5char4_Z6__ldluPK6short2_Z6__ldluPK6short4_Z6__ldluPK4int2_Z6__ldluPK4int4_Z6__ldluPK9longlong2_Z6__ldluPKh_Z6__ldluPKt_Z6__ldluPKj_Z6__ldluPKy_Z6__ldluPK6uchar2_Z6__ldluPK6uchar4_Z6__ldluPK7ushort2_Z6__ldluPK7ushort4_Z6__ldluPK5uint2_Z6__ldluPK5uint4_Z6__ldluPK10ulonglong2_Z6__ldluPKf_Z6__ldluPKd_Z6__ldluPK6float2_Z6__ldluPK6float4_Z6__ldluPK7double2_Z6__ldcvPKl_Z6__ldcvPKm_Z6__ldcvPKc_Z6__ldcvPKa_Z6__ldcvPKs_Z6__ldcvPKi_Z6__ldcvPKx_Z6__ldcvPK5char2_Z6__ldcvPK5char4_Z6__ldcvPK6short2_Z6__ldcvPK6short4_Z6__ldcvPK4int2_Z6__ldcvPK4int4_Z6__ldcvPK9longlong2_Z6__ldcvPKh_Z6__ldcvPKt_Z6__ldcvPKj_Z6__ldcvPKy_Z6__ldcvPK6uchar2_Z6__ldcvPK6uchar4_Z6__ldcvPK7ushort2_Z6__ldcvPK7ushort4_Z6__ldcvPK5uint2_Z6__ldcvPK5uint4_Z6__ldcvPK10ulonglong2_Z6__ldcvPKf_Z6__ldcvPKd_Z6__ldcvPK6float2_Z6__ldcvPK6float4_Z6__ldcvPK7double2_Z6__stwbPll_Z6__stwbPmm_Z6__stwbPcc_Z6__stwbPaa_Z6__stwbPss_Z6__stwbPii_Z6__stwbPxx_Z6__stwbP5char2S__Z6__stwbP5char4S__Z6__stwbP6short2S__Z6__stwbP6short4S__Z6__stwbP4int2S__Z6__stwbP4int4S__Z6__stwbP9longlong2S__Z6__stwbPhh_Z6__stwbPtt_Z6__stwbPjj_Z6__stwbPyy_Z6__stwbP6uchar2S__Z6__stwbP6uchar4S__Z6__stwbP7ushort2S__Z6__stwbP7ushort4S__Z6__stwbP5uint2S__Z6__stwbP5uint4S__Z6__stwbP10ulonglong2S__Z6__stwbPff_Z6__stwbPdd_Z6__stwbP6float2S__Z6__stwbP6float4S__Z6__stwbP7double2S__Z6__stcgPll_Z6__stcgPmm_Z6__stcgPcc_Z6__stcgPaa_Z6__stcgPss_Z6__stcgPii_Z6__stcgPxx_Z6__stcgP5char2S__Z6__stcgP5char4S__Z6__stcgP6short2S__Z6__stcgP6short4S__Z6__stcgP4int2S__Z6__stcgP4int4S__Z6__stcgP9longlong2S__Z6__stcgPhh_Z6__stcgPtt_Z6__stcgPjj_Z6__stcgPyy_Z6__stcgP6uchar2S__Z6__stcgP6uchar4S__Z6__stcgP7ushort2S__Z6__stcgP7ushort4S__Z6__stcgP5uint2S__Z6__stcgP5uint4S__Z6__stcgP10ulonglong2S__Z6__stcgPff_Z6__stcgPdd_Z6__stcgP6float2S__Z6__stcgP6float4S__Z6__stcgP7double2S__Z6__stcsPll_Z6__stcsPmm_Z6__stcsPcc_Z6__stcsPaa_Z6__stcsPss_Z6__stcsPii_Z6__stcsPxx_Z6__stcsP5char2S__Z6__stcsP5char4S__Z6__stcsP6short2S__Z6__stcsP6short4S__Z6__stcsP4int2S__Z6__stcsP4int4S__Z6__stcsP9longlong2S__Z6__stcsPhh_Z6__stcsPtt_Z6__stcsPjj_Z6__stcsPyy_Z6__stcsP6uchar2S__Z6__stcsP6uchar4S__Z6__stcsP7ushort2S__Z6__stcsP7ushort4S__Z6__stcsP5uint2S__Z6__stcsP5uint4S__Z6__stcsP10ulonglong2S__Z6__stcsPff_Z6__stcsPdd_Z6__stcsP6float2S__Z6__stcsP6float4S__Z6__stcsP7double2S__Z6__stwtPll_Z6__stwtPmm_Z6__stwtPcc_Z6__stwtPaa_Z6__stwtPss_Z6__stwtPii_Z6__stwtPxx_Z6__stwtP5char2S__Z6__stwtP5char4S__Z6__stwtP6short2S__Z6__stwtP6short4S__Z6__stwtP4int2S__Z6__stwtP4int4S__Z6__stwtP9longlong2S__Z6__stwtPhh_Z6__stwtPtt_Z6__stwtPjj_Z6__stwtPyy_Z6__stwtP6uchar2S__Z6__stwtP6uchar4S__Z6__stwtP7ushort2S__Z6__stwtP7ushort4S__Z6__stwtP5uint2S__Z6__stwtP5uint4S__Z6__stwtP10ulonglong2S__Z6__stwtPff_Z6__stwtPdd_Z6__stwtP6float2S__Z6__stwtP6float4S__Z6__stwtP7double2S__Z15__funnelshift_ljjj_Z16__funnelshift_lcjjj_Z15__funnelshift_rjjj_Z16__funnelshift_rcjjj_Z16__match_any_syncjjllvm.nvvm.match.any.sync.i32_Z16__match_any_syncji_Z16__match_any_syncjmllvm.nvvm.match.any.sync.i64_Z16__match_any_syncjl_Z16__match_any_syncjy_Z16__match_any_syncjx_Z16__match_any_syncjf__float_as_uint_Z16__match_any_syncjd__double_as_longlong_Z16__match_all_syncjjPillvm.nvvm.match.all.sync.i32_Z16__match_all_syncjiPi_Z16__match_all_syncjmPillvm.nvvm.match.all.sync.i64_Z16__match_all_syncjlPi_Z16__match_all_syncjyPi_Z16__match_all_syncjxPi_Z16__match_all_syncjfPi_Z16__match_all_syncjdPi_Z11__nanosleepj_Z9atomicCASPttt__usAtomicCAS_Z17__reduce_add_syncjj__reduce_add_sync_unsigned_impl_Z17__reduce_min_syncjj__reduce_min_sync_unsigned_impl_Z17__reduce_max_syncjj__reduce_max_sync_unsigned_impl_Z17__reduce_add_syncji__reduce_add_sync_signed_impl_Z17__reduce_min_syncji__reduce_min_sync_signed_impl_Z17__reduce_max_syncji__reduce_max_sync_signed_impl_Z17__reduce_and_syncjj__reduce_and_sync_unsigned_impl_Z16__reduce_or_syncjj__reduce_or_sync_unsigned_impl_Z17__reduce_xor_syncjj__reduce_xor_sync_unsigned_impl_Z13__isCtaSharedPKv_Z17__isClusterSharedPKvllvm.nvvm.isspacep.cluster.shared_Z25__cluster_map_shared_rankPKvjllvm.nvvm.cluster.set.rank_Z27__cluster_query_shared_rankPKvllvm.nvvm.cluster.get.rank_Z30__cluster_map_shared_multicastPKvj_Z23__clusterDimIsSpecifiedvllvm.nvvm.read.cluster.info.i1_Z12__clusterDimvllvm.nvvm.read.cluster.info.i32_Z25__clusterRelativeBlockIdxv_Z26__clusterGridDimInClustersv_Z12__clusterIdxv_Z26__clusterRelativeBlockRankv_Z21__clusterSizeInBlocksv_Z24__cluster_barrier_arrivevllvm.nvvm.cluster.barrier_Z32__cluster_barrier_arrive_relaxedv_Z22__cluster_barrier_waitv_Z21__threadfence_clustervllvm.nvvm.membar_Z9atomicAddP6float2S___f2AtomicAdd_Z15atomicAdd_blockP6float2S___f2AtomicAdd_block_Z16atomicAdd_systemP6float2S___f2AtomicAdd_system_Z9atomicAddP6float4S___f4AtomicAdd_Z15atomicAdd_blockP6float4S___f4AtomicAdd_block_Z16atomicAdd_systemP6float4S___f4AtomicAdd_system7.0.1nvptx64-nvidia-gpulibsmoduleOutput;��̲��ܲ��첗��,��D��\,��t<��L��\��l��|��4��L̳��d��|zRx�$谗�FJw�?;*3$"D౗�\ر��tб��ȱ��4��L��d��|x��p��h��`�� ` �]��'� h �]��]��o�� M`�� ox��o��o��o��oF��o�]�.shstrtab.note.gnu.build-id.gnu.hash.dynsym.dynstr.gnu.version.gnu.version_d.gnu.version_r.rela.dyn.init.plt.got.text.fini.rodata.eh_frame_hdr.eh_frame.init_array.fini_array.jcr.data.rel.ro.dynamic.got.plt.bss��$��o��(��X0��M8��oFF2E��oxx0T��o�� c��m� � �� s� � |� � ��h h �� Jh �XhXh��Xh�Xh��]��]h��]��]h��]��]h��]��]h��]��]hw�_��_h(�`�`h�`�`h`h�