Add hipDrvOccupancyMaxActiveBlocksPerMultiprocessor[WithFlags] (#1854)

Equivalent to cuOccupancyMaxActiveBlocksPerMultiprocessor[WithFlags].
Tento commit je obsažen v:
Rahul Garg
2020-02-28 03:16:55 -08:00
odevzdal GitHub
rodič e747efe2b4
revize 1c794045e0
4 změnil soubory, kde provedl 48 přidání a 7 odebrání
+24 -1
Zobrazit soubor
@@ -2959,7 +2959,18 @@ hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(
* @brief Returns occupancy for a device function.
*
* @param [out] numBlocks Returned occupancy
* @param [in] func Kernel function for which occupancy is calulated
* @param [in] func Kernel function (hipFunction) 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
*/
hipError_t hipDrvOccupancyMaxActiveBlocksPerMultiprocessor(
int* numBlocks, hipFunction_t f, int blockSize, size_t dynSharedMemPerBlk);
/**
* @brief Returns occupancy for a device function.
*
* @param [out] numBlocks Returned occupancy
* @param [in] f Kernel function 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)
@@ -2967,6 +2978,18 @@ hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessor(
hipError_t hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
uint32_t* numBlocks, hipFunction_t f, uint32_t blockSize, size_t dynSharedMemPerBlk, unsigned int flags __dparm(hipOccupancyDefault));
/**
* @brief Returns occupancy for a device function.
*
* @param [out] numBlocks Returned occupancy
* @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)
*/
hipError_t hipDrvOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(
int* numBlocks, hipFunction_t f, int blockSize, size_t dynSharedMemPerBlk, unsigned int flags);
#if __HIP_VDI__ && !defined(__HCC__)
/**
* @brief Launches kernels on multiple devices and guarantees all specified kernels are dispatched