Rename hipDrvOccupancy to hipModuleOccupancy and match CUDA syntax (#1943)
This commit is contained in:
Rahul Garg
GitHub
@@ -140,10 +140,10 @@ void hipLaunchKernelGGLImpl(
|
||||
} // Namespace hip_impl.
|
||||
|
||||
|
||||
template <typename F>
|
||||
template <class T>
|
||||
inline
|
||||
hipError_t hipOccupancyMaxPotentialBlockSize(uint32_t* gridSize, uint32_t* blockSize,
|
||||
F kernel, size_t dynSharedMemPerBlk, uint32_t blockSizeLimit) {
|
||||
hipError_t hipOccupancyMaxPotentialBlockSize(int* gridSize, int* blockSize,
|
||||
T kernel, size_t dynSharedMemPerBlk = 0, int blockSizeLimit = 0) {
|
||||
|
||||
using namespace hip_impl;
|
||||
|
||||
@@ -151,22 +151,24 @@ hipError_t hipOccupancyMaxPotentialBlockSize(uint32_t* gridSize, uint32_t* block
|
||||
auto f = get_program_state().kernel_descriptor(reinterpret_cast<std::uintptr_t>(kernel),
|
||||
target_agent(0));
|
||||
|
||||
return hipOccupancyMaxPotentialBlockSize(gridSize, blockSize, f,
|
||||
return hipModuleOccupancyMaxPotentialBlockSize(gridSize, blockSize, f,
|
||||
dynSharedMemPerBlk, blockSizeLimit);
|
||||
}
|
||||
|
||||
template <typename F>
|
||||
template <class T>
|
||||
inline
|
||||
hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(uint32_t* numBlocks, F kernel,
|
||||
uint32_t blockSize, size_t dynSharedMemPerBlk) {
|
||||
hipError_t hipOccupancyMaxPotentialBlockSizeWithFlags(int* gridSize, int* blockSize,
|
||||
T kernel, size_t dynSharedMemPerBlk = 0, int blockSizeLimit = 0, unsigned int flags = 0 ) {
|
||||
|
||||
using namespace hip_impl;
|
||||
|
||||
hip_impl::hip_init();
|
||||
if(flags != hipOccupancyDefault) return hipErrorNotSupported;
|
||||
auto f = get_program_state().kernel_descriptor(reinterpret_cast<std::uintptr_t>(kernel),
|
||||
target_agent(0));
|
||||
|
||||
return hipOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, f, blockSize, dynSharedMemPerBlk);
|
||||
return hipModuleOccupancyMaxPotentialBlockSize(gridSize, blockSize, f,
|
||||
dynSharedMemPerBlk, blockSizeLimit);
|
||||
}
|
||||
|
||||
template <typename... Args, typename F = void (*)(Args...)>
|
||||
|
||||
@@ -266,7 +266,6 @@ typedef enum hipSharedMemConfig {
|
||||
///< when adjacent threads access data 4 bytes apart.
|
||||
} hipSharedMemConfig;
|
||||
|
||||
|
||||
/**
|
||||
* Struct for data in 3D
|
||||
*
|
||||
@@ -2940,9 +2939,28 @@ hipError_t hipLaunchCooperativeKernelMultiDevice(hipLaunchParams* launchParamsLi
|
||||
*
|
||||
* @returns hipSuccess, hipInvalidDevice, hipErrorInvalidValue
|
||||
*/
|
||||
hipError_t hipOccupancyMaxPotentialBlockSize(uint32_t* gridSize, uint32_t* blockSize,
|
||||
|
||||
//TODO - Match CUoccupancyB2DSize
|
||||
hipError_t hipModuleOccupancyMaxPotentialBlockSize(int* gridSize, int* blockSize,
|
||||
hipFunction_t f, size_t dynSharedMemPerBlk,
|
||||
uint32_t blockSizeLimit);
|
||||
int blockSizeLimit);
|
||||
|
||||
/**
|
||||
* @brief determine the grid and block sizes to achieves maximum occupancy for a kernel
|
||||
*
|
||||
* @param [out] gridSize minimum grid size for maximum potential occupancy
|
||||
* @param [out] blockSize block size for maximum potential occupancy
|
||||
* @param [in] f kernel function for which occupancy is calulated
|
||||
* @param [in] dynSharedMemPerBlk dynamic shared memory usage (in bytes) intended for each block
|
||||
* @param [in] blockSizeLimit the maximum block size for the kernel, use 0 for no limit
|
||||
* @param [in] flags Extra flags for occupancy calculation (only default supported)
|
||||
*
|
||||
* @returns hipSuccess, hipInvalidDevice, hipErrorInvalidValue
|
||||
*/
|
||||
//TODO - Match CUoccupancyB2DSize
|
||||
hipError_t hipModuleOccupancyMaxPotentialBlockSizeWithFlags(int* gridSize, int* blockSize,
|
||||
hipFunction_t f, size_t dynSharedMemPerBlk,
|
||||
int blockSizeLimit, unsigned int flags);
|
||||
|
||||
/**
|
||||
* @brief Returns occupancy for a device function.
|
||||
@@ -2953,7 +2971,7 @@ hipError_t hipOccupancyMaxPotentialBlockSize(uint32_t* gridSize, uint32_t* block
|
||||
* @param [in] dynSharedMemPerBlk dynamic shared memory usage (in bytes) intended for each block
|
||||
*/
|
||||
hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
uint32_t* numBlocks, hipFunction_t f, uint32_t blockSize, size_t dynSharedMemPerBlk);
|
||||
int* numBlocks, const void* f, int blockSize, size_t dynSharedMemPerBlk);
|
||||
|
||||
/**
|
||||
* @brief Returns occupancy for a device function.
|
||||
@@ -2963,7 +2981,7 @@ hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
* @param [in] blockSize Block size the kernel is intended to be launched with
|
||||
* @param [in] dynSharedMemPerBlk dynamic shared memory usage (in bytes) intended for each block
|
||||
*/
|
||||
hipError_t hipDrvOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
hipError_t hipModuleOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
int* numBlocks, hipFunction_t f, int blockSize, size_t dynSharedMemPerBlk);
|
||||
|
||||
/**
|
||||
@@ -2976,7 +2994,7 @@ hipError_t hipDrvOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
* @param [in] flags Extra flags for occupancy calculation (currently ignored)
|
||||
*/
|
||||
hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
uint32_t* numBlocks, hipFunction_t f, uint32_t blockSize, size_t dynSharedMemPerBlk, unsigned int flags __dparm(hipOccupancyDefault));
|
||||
int* numBlocks, const void* f, int blockSize, size_t dynSharedMemPerBlk, unsigned int flags __dparm(hipOccupancyDefault));
|
||||
|
||||
/**
|
||||
* @brief Returns occupancy for a device function.
|
||||
@@ -2985,9 +3003,9 @@ hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
* @param [in] f Kernel function(hipFunction_t) for which occupancy is calulated
|
||||
* @param [in] blockSize Block size the kernel is intended to be launched with
|
||||
* @param [in] dynSharedMemPerBlk dynamic shared memory usage (in bytes) intended for each block
|
||||
* @param [in] flags Extra flags for occupancy calculation (currently ignored)
|
||||
* @param [in] flags Extra flags for occupancy calculation (only default supported)
|
||||
*/
|
||||
hipError_t hipDrvOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
hipError_t hipModuleOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
int* numBlocks, hipFunction_t f, int blockSize, size_t dynSharedMemPerBlk, unsigned int flags);
|
||||
|
||||
#if __HIP_VDI__ && !defined(__HCC__)
|
||||
@@ -3258,21 +3276,6 @@ hipError_t hipLaunchKernel(const void* function_address,
|
||||
} /* extern "c" */
|
||||
#endif
|
||||
|
||||
#if defined(__cplusplus) && !defined(__HCC__) && defined(__clang__) && defined(__HIP__)
|
||||
template <typename F>
|
||||
static hipError_t __host__ inline hipOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
uint32_t* numBlocks, F func, uint32_t blockSize, size_t dynSharedMemPerBlk) {
|
||||
return ::hipOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, (hipFunction_t)func, blockSize,
|
||||
dynSharedMemPerBlk);
|
||||
}
|
||||
template <typename F>
|
||||
static hipError_t __host__ inline hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
uint32_t* numBlocks, F func, uint32_t blockSize, size_t dynSharedMemPerBlk, unsigned int flags) {
|
||||
return ::hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
numBlocks, (hipFunction_t)func, blockSize, dynSharedMemPerBlk, flags);
|
||||
}
|
||||
#endif // defined(__cplusplus) && !defined(__HCC__) && defined(__clang__) && defined(__HIP__)
|
||||
|
||||
#if USE_PROF_API
|
||||
#include <hip/hcc_detail/hip_prof_str.h>
|
||||
#endif
|
||||
@@ -3295,6 +3298,20 @@ const char* hipKernelNameRef(const hipFunction_t f);
|
||||
|
||||
#ifdef __cplusplus
|
||||
|
||||
template <class T>
|
||||
inline hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
int* numBlocks, T f, int blockSize, size_t dynSharedMemPerBlk) {
|
||||
return hipOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
numBlocks, reinterpret_cast<const void*>(f), blockSize, dynSharedMemPerBlk);
|
||||
}
|
||||
|
||||
template <class T>
|
||||
inline hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
int* numBlocks, T f, int blockSize, size_t dynSharedMemPerBlk, unsigned int flags) {
|
||||
return hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
|
||||
numBlocks, reinterpret_cast<const void*>(f), blockSize, dynSharedMemPerBlk, flags);
|
||||
}
|
||||
|
||||
class TlsData;
|
||||
|
||||
hipError_t hipBindTexture(size_t* offset, textureReference* tex, const void* devPtr,
|
||||
|
||||
@@ -1295,10 +1295,50 @@ inline static hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(int* numBl
|
||||
const void* func,
|
||||
int blockSize,
|
||||
size_t dynamicSMemSize) {
|
||||
cudaError_t cerror;
|
||||
cerror =
|
||||
cudaOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSMemSize);
|
||||
return hipCUDAErrorTohipError(cerror);
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func,
|
||||
blockSize, dynamicSMemSize));
|
||||
}
|
||||
|
||||
inline static hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int* numBlocks,
|
||||
const void* func,
|
||||
int blockSize,
|
||||
size_t dynamicSMemSize,
|
||||
unsigned int flags) {
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func,
|
||||
blockSize, dynamicSMemSize, flags));
|
||||
}
|
||||
|
||||
inline static hipError_t hipModuleOccupancyMaxActiveBlocksPerMultiprocessor(int* numBlocks,
|
||||
hipFunction_t f,
|
||||
int blockSize,
|
||||
size_t dynamicSMemSize ){
|
||||
return hipCUResultTohipError(cuOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, f,
|
||||
blockSize, dynamicSMemSize));
|
||||
}
|
||||
|
||||
inline static hipError_t hipModuleOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int* numBlocks,
|
||||
hipFunction_t f,
|
||||
int blockSize,
|
||||
size_t dynamicSMemSize,
|
||||
unsigned int flags ) {
|
||||
return hipCUResultTohipError(cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks,f,
|
||||
blockSize, dynamicSMemSize, flags));
|
||||
}
|
||||
|
||||
//TODO - Match CUoccupancyB2DSize
|
||||
inline static hipError_t hipModuleOccupancyMaxPotentialBlockSize(int* gridSize, int* blockSize,
|
||||
hipFunction_t f, size_t dynSharedMemPerBlk,
|
||||
int blockSizeLimit){
|
||||
return hipCUResultTohipError(cuOccupancyMaxPotentialBlockSize(gridSize, blockSize, f, NULL,
|
||||
dynSharedMemPerBlk, blockSizeLimit));
|
||||
}
|
||||
|
||||
//TODO - Match CUoccupancyB2DSize
|
||||
inline static hipError_t hipModuleOccupancyMaxPotentialBlockSizeWithFlags(int* gridSize, int* blockSize,
|
||||
hipFunction_t f, size_t dynSharedMemPerBlk,
|
||||
int blockSizeLimit, unsigned int flags){
|
||||
return hipCUResultTohipError(cuOccupancyMaxPotentialBlockSizeWithFlags(gridSize, blockSize, f, NULL,
|
||||
dynSharedMemPerBlk, blockSizeLimit, flags));
|
||||
}
|
||||
|
||||
inline static hipError_t hipPointerGetAttributes(hipPointerAttribute_t* attributes, const void* ptr) {
|
||||
@@ -1713,19 +1753,31 @@ inline static hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(int* numBl
|
||||
T func,
|
||||
int blockSize,
|
||||
size_t dynamicSMemSize) {
|
||||
cudaError_t cerror;
|
||||
cerror =
|
||||
cudaOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func, blockSize, dynamicSMemSize);
|
||||
return hipCUDAErrorTohipError(cerror);
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxActiveBlocksPerMultiprocessor(numBlocks, func,
|
||||
blockSize, dynamicSMemSize));
|
||||
}
|
||||
|
||||
template <class T>
|
||||
inline static hipError_t hipOccupancyMaxPotentialBlockSize(int* minGridSize, int* blockSize, T func,
|
||||
size_t dynamicSMemSize = 0,
|
||||
int blockSizeLimit = 0) {
|
||||
cudaError_t cerror;
|
||||
cerror = cudaOccupancyMaxPotentialBlockSize(minGridSize, blockSize, func, dynamicSMemSize, blockSizeLimit);
|
||||
return hipCUDAErrorTohipError(cerror);
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxPotentialBlockSize(minGridSize, blockSize, func,
|
||||
dynamicSMemSize, blockSizeLimit));
|
||||
}
|
||||
|
||||
template <class T>
|
||||
inline static hipError_t hipOccupancyMaxPotentialBlockSizeWithFlags(int* minGridSize, int* blockSize, T func,
|
||||
size_t dynamicSMemSize = 0,
|
||||
int blockSizeLimit = 0, unsigned int flags = 0) {
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxPotentialBlockSize(minGridSize, blockSize, func,
|
||||
dynamicSMemSize, blockSizeLimit, flags));
|
||||
}
|
||||
|
||||
template <class T>
|
||||
inline static hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( int* numBlocks, T func,
|
||||
int blockSize, size_t dynamicSMemSize,unsigned int flags) {
|
||||
return hipCUDAErrorTohipError(cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(numBlocks, func,
|
||||
blockSize, dynamicSMemSize, flags));
|
||||
}
|
||||
|
||||
template <class T, int dim, enum cudaTextureReadMode readMode>
|
||||
|
||||
مرجع در شماره جدید
Block a user