CUDA 10.0 隐藏的 Runtime API 和 Driver API（__ 开头）

以下是 CUDA 10.0 中隐藏的一些 Runtime API 和 Driver API，它们以 '__' 开头：

1. Runtime API:

• '__cudaRegisterFunction()':

cudaError_t __cudaRegisterFunction(void **fatCubinHandle, const char *hostFun, char *deviceFun, const char *deviceName, int thread_limit, uint3 *tid, uint3 *bid, dim3 *bDim, dim3 *gDim, int *wSize);

• '__cudaRegisterVar()':

cudaError_t __cudaRegisterVar(void **fatCubinHandle, char *hostVar, char *deviceAddress, const char *deviceName, int ext, size_t size, int constant, int global);

• '__cudaRegisterManagedVar()':

cudaError_t __cudaRegisterManagedVar(void **fatCubinHandle, void *hostVar, size_t size, int constant, int global);

• '__cudaRegisterTexture()':

cudaError_t __cudaRegisterTexture(struct textureReference *hostVar, const void **deviceAddress, const char *deviceName, int dim, int norm, int ext);

• '__cudaRegisterSurface()':

cudaError_t __cudaRegisterSurface(struct surfaceReference *hostVar, const void **deviceAddress, const char *deviceName, int dim, int ext);

• '__cudaRegisterFunctionTexture()':

cudaError_t __cudaRegisterFunctionTexture(void **fatCubinHandle, const char *hostFun, char *deviceFun, const char *deviceName, int thread_limit, uint3 *tid, uint3 *bid, dim3 *bDim, dim3 *gDim, int *wSize, struct textureReference *hostVar, const void **deviceAddress);

• '__cudaRegisterFunctionSurface()':

cudaError_t __cudaRegisterFunctionSurface(void **fatCubinHandle, const char *hostFun, char *deviceFun, const char *deviceName, int thread_limit, uint3 *tid, uint3 *bid, dim3 *bDim, dim3 *gDim, int *wSize, struct surfaceReference *hostVar, const void **deviceAddress);

• '__cudaRegisterFatBinary()':

cudaError_t __cudaRegisterFatBinary(void **fatCubinHandle);

• '__cudaRegisterFatBinaryEnd()':

void __cudaRegisterFatBinaryEnd(void **fatCubinHandle);

• '__cudaUnregisterFatBinary()':

cudaError_t __cudaUnregisterFatBinary(void **fatCubinHandle);

• '__cudaPopCallConfiguration()':

cudaError_t __cudaPopCallConfiguration(dim3 *gridDim, dim3 *blockDim, size_t *sharedMem, void *stream);

• '__cudaPushCallConfiguration()':

cudaError_t __cudaPushCallConfiguration(dim3 gridDim, dim3 blockDim, size_t sharedMem, void *stream);

2. Driver API:

• '__cudaRegisterFatBinary()':

CUresult __cudaRegisterFatBinary(void *fatCubin);

• '__cudaRegisterFatBinaryEnd()':

CUresult __cudaRegisterFatBinaryEnd(void);

• '__cudaUnregisterFatBinary()':

CUresult __cudaUnregisterFatBinary(void *fatCubin);

• '__cudaPopCallConfiguration()':

CUresult __cudaPopCallConfiguration(CUDA_POP_CALL_ARGS);

• '__cudaPushCallConfiguration()':

CUresult __cudaPushCallConfiguration(CUDA_PUSH_CALL_ARGS);