#include "mlir/ExecutionEngine/CRunnerUtils.h"
#include <cstdio>
#include "cuda.h"
#include "cuda_bf16.h"
#include "cuda_fp16.h"

Classes
class	ScopedContext

Macros
#define	MLIR_CUDA_WRAPPERS_EXPORT __attribute__((visibility("default")))

#define	CUDA_REPORT_IF_ERROR(expr)

#define	CUSPARSE_REPORT_IF_ERROR(expr)

#define	debug_print(fmt, ...)

Functions
bool	isDebugEnabled ()
	Helper method that checks environment value for debugging. More...

CUdevice	getDefaultCuDevice ()

MLIR_CUDA_WRAPPERS_EXPORT CUmodule	mgpuModuleLoad (void *data, size_t)

MLIR_CUDA_WRAPPERS_EXPORT CUmodule	mgpuModuleLoadJIT (void *data, int optLevel)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuModuleUnload (CUmodule module)

MLIR_CUDA_WRAPPERS_EXPORT CUfunction	mgpuModuleGetFunction (CUmodule module, const char *name)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuLaunchKernel (CUfunction function, intptr_t gridX, intptr_t gridY, intptr_t gridZ, intptr_t blockX, intptr_t blockY, intptr_t blockZ, int32_t smem, CUstream stream, void params, void extra, size_t)

MLIR_CUDA_WRAPPERS_EXPORT CUstream	mgpuStreamCreate ()

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuStreamDestroy (CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuStreamSynchronize (CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuStreamWaitEvent (CUstream stream, CUevent event)

MLIR_CUDA_WRAPPERS_EXPORT CUevent	mgpuEventCreate ()

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuEventDestroy (CUevent event)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuEventSynchronize (CUevent event)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuEventRecord (CUevent event, CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void *	mgpuMemAlloc (uint64_t sizeBytes, CUstream stream, bool isHostShared)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemFree (void *ptr, CUstream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemcpy (void dst, void src, size_t sizeBytes, CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemset32 (void *dst, unsigned int value, size_t count, CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemset16 (void *dst, unsigned short value, size_t count, CUstream stream)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemHostRegister (void *ptr, uint64_t sizeBytes)
	Helper functions for writing mlir example code. More...

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemHostRegisterMemRef (int64_t rank, StridedMemRefType< char, 1 > *descriptor, int64_t elementSizeBytes)
	Registers a memref with the CUDA runtime. More...

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemHostUnregister (void *ptr)

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuMemHostUnregisterMemRef (int64_t rank, StridedMemRefType< char, 1 > *descriptor, int64_t elementSizeBytes)
	Unregisters a memref with the CUDA runtime. More...

MLIR_CUDA_WRAPPERS_EXPORT void	mgpuSetDefaultDevice (int32_t device)

Variables
static thread_local int32_t	defaultDevice = 0

Macro Definition Documentation

◆ CUDA_REPORT_IF_ERROR

#define CUDA_REPORT_IF_ERROR ( expr )

Value:

  [](CUresult result) {                                                        \
    if (!result)                                                               \
      return;                                                                  \
    const char *name = nullptr;                                                \
    cuGetErrorName(result, &name);                                             \
    if (!name)                                                                 \
      name = "<unknown>";                                                      \
    fprintf(stderr, "'%s' failed with '%s'\n", #expr, name);                   \
  }(expr)

Definition at line 37 of file CudaRuntimeWrappers.cpp.

◆ CUSPARSE_REPORT_IF_ERROR

#define CUSPARSE_REPORT_IF_ERROR ( expr )

Value:

  {                                                                            \
    cusparseStatus_t status = (expr);                                          \
    if (status != CUSPARSE_STATUS_SUCCESS) {                                   \
      fprintf(stderr, "cuSPARSE '%s' failed with '%s'\n", #expr,               \
              cusparseGetErrorString(status));                                 \
    }                                                                          \
  }

Definition at line 48 of file CudaRuntimeWrappers.cpp.

◆ debug_print

#define debug_print	(	fmt,
		...
	)

Value:

  do {                                                                         \
    if (isDebugEnabled())                                                      \
      fprintf(stderr, "%s:%d:%s(): " fmt, "CudaRuntimeWrappers.cpp", __LINE__, \
              __func__, __VA_ARGS__);                                          \
  } while (0)

Definition at line 66 of file CudaRuntimeWrappers.cpp.

◆ MLIR_CUDA_WRAPPERS_EXPORT

#define MLIR_CUDA_WRAPPERS_EXPORT __attribute__((visibility("default")))

Definition at line 34 of file CudaRuntimeWrappers.cpp.

Function Documentation

◆ getDefaultCuDevice()

CUdevice getDefaultCuDevice ( )

Definition at line 74 of file CudaRuntimeWrappers.cpp.

References CUDA_REPORT_IF_ERROR, and defaultDevice.

Referenced by mgpuLaunchKernel(), and ScopedContext::ScopedContext().

◆ isDebugEnabled()

bool isDebugEnabled ( )

Helper method that checks environment value for debugging.

Definition at line 60 of file CudaRuntimeWrappers.cpp.

◆ mgpuEventCreate()

MLIR_CUDA_WRAPPERS_EXPORT CUevent mgpuEventCreate ( )

Definition at line 215 of file CudaRuntimeWrappers.cpp.

References CUDA_REPORT_IF_ERROR.

◆ mgpuEventDestroy()

MLIR_CUDA_WRAPPERS_EXPORT void mgpuEventDestroy ( CUevent event )

Definition at line 222 of file CudaRuntimeWrappers.cpp.

References CUDA_REPORT_IF_ERROR.

◆ mgpuEventRecord()

MLIR_CUDA_WRAPPERS_EXPORT void mgpuEventRecord	(	CUevent	event,
		CUstream	stream
	)

Definition at line 230 of file CudaRuntimeWrappers.cpp.

References CUDA_REPORT_IF_ERROR.

◆ mgpuEventSynchronize()

MLIR_CUDA_WRAPPERS_EXPORT void mgpuEventSynchronize ( CUevent event )

Definition at line 226 of file CudaRuntimeWrappers.cpp.

References CUDA_REPORT_IF_ERROR.

◆ mgpuLaunchKernel()

MLIR_CUDA_WRAPPERS_EXPORT void mgpuLaunchKernel	(	CUfunction	function,
		intptr_t	gridX,
		intptr_t	gridY,
		intptr_t	gridZ,
		intptr_t	blockX,
		intptr_t	blockY,
		intptr_t	blockZ,
		int32_t	smem,
		CUstream	stream,
		void **	params,
		void **	extra,
		size_t
	)