include/hip/hip_ext.h Source File

include/hip/hip_ext.h Source File#

HIP Runtime API Reference: include/hip/hip_ext.h Source File
/*
 * Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
 *
 * SPDX-License-Identifier: MIT
 */
 
#ifndef HIP_INCLUDE_HIP_HIP_EXT_H
#define HIP_INCLUDE_HIP_HIP_EXT_H
#include "hip/hip_runtime.h"
#if defined(__cplusplus)
#include <tuple>
#include <type_traits>
#endif
HIP_PUBLIC_API
extern "C" hipError_t hipExtModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
                                               uint32_t globalWorkSizeY, uint32_t globalWorkSizeZ,
                                               uint32_t localWorkSizeX, uint32_t localWorkSizeY,
                                               uint32_t localWorkSizeZ, size_t sharedMemBytes,
                                               hipStream_t hStream, void** kernelParams,
                                               void** extra, hipEvent_t startEvent __dparm(NULL),
                                               hipEvent_t stopEvent __dparm(NULL),
                                               uint32_t flags __dparm(0));
HIP_DEPRECATED("use hipExtModuleLaunchKernel instead")
HIP_PUBLIC_API
extern "C" hipError_t hipHccModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
                                               uint32_t globalWorkSizeY, uint32_t globalWorkSizeZ,
                                               uint32_t localWorkSizeX, uint32_t localWorkSizeY,
                                               uint32_t localWorkSizeZ, size_t sharedMemBytes,
                                               hipStream_t hStream, void** kernelParams,
                                               void** extra, hipEvent_t startEvent __dparm(NULL),
                                               hipEvent_t stopEvent __dparm(NULL));
 
#if defined(__cplusplus)
 
extern "C" hipError_t hipExtLaunchKernel(const void* function_address, dim3 numBlocks,
                                         dim3 dimBlocks, void** args, size_t sharedMemBytes,
                                         hipStream_t stream, hipEvent_t startEvent,
                                         hipEvent_t stopEvent, int flags);
 
template <typename... Args, typename F = void (*)(Args...)>
inline void hipExtLaunchKernelGGL(F kernel, const dim3& numBlocks, const dim3& dimBlocks,
                                  std::uint32_t sharedMemBytes, hipStream_t stream,
                                  hipEvent_t startEvent, hipEvent_t stopEvent, std::uint32_t flags,
                                  Args... args) {
  constexpr size_t count = sizeof...(Args);
  auto tup_ = std::tuple<Args...>{args...};
  auto tup = validateArgsCountType(kernel, tup_);
  void* _Args[count];
  pArgs<0>(tup, _Args);
 
  auto k = reinterpret_cast<void*>(kernel);
  hipExtLaunchKernel(k, numBlocks, dimBlocks, _Args, sharedMemBytes, stream, startEvent, stopEvent,
                     (int)flags);
}
 
#endif  // defined(__cplusplus)
 
// doxygen end AMD-specific features
#endif  // #iidef HIP_INCLUDE_HIP_HIP_EXT_H