quda_cuda_api.h

Go to the documentation of this file.

#pragma once

#include <cuda.h>
#include <cuda_runtime.h>
#include <quda_cuda_api.h>

namespace quda {

  void qudaMemcpy_(void *dst, const void *src, size_t count, cudaMemcpyKind kind,
       const char *func, const char *file, const char *line);

}

#define STRINGIFY__(x) #x
#define __STRINGIFY__(x) STRINGIFY__(x)
#define qudaMemcpy(dst, src, count, kind) \
  ::quda::qudaMemcpy_(dst, src, count, kind, __func__, quda::file_name(__FILE__), __STRINGIFY__(__LINE__));

#define STRINGIFY__(x) #x
#define __STRINGIFY__(x) STRINGIFY__(x)
#define qudaMemcpyAsync(dst, src, count, kind, stream) \
  ::quda::qudaMemcpyAsync_(dst, src, count, kind, stream, __func__, quda::file_name(__FILE__), __STRINGIFY__(__LINE__));

#define STRINGIFY__(x) #x
#define __STRINGIFY__(x) STRINGIFY__(x)
#define qudaMemcpy2DAsync(dst, dpitch, src, spitch, width, height, kind, stream) \
  ::quda::qudaMemcpy2DAsync_(dst, dpitch, src, spitch, width, height, kind, stream, __func__, quda::file_name(__FILE__), __STRINGIFY__(__LINE__));

namespace quda {

  void qudaMemcpyAsync_(void *dst, const void *src, size_t count, cudaMemcpyKind kind, const cudaStream_t &stream,
                        const char *func, const char *file, const char *line);

  void qudaMemcpy2DAsync_(void *dst, size_t dpitch, const void *src, size_t spitch,
                          size_t width, size_t hieght, cudaMemcpyKind kind, const cudaStream_t &stream,
                          const char *func, const char *file, const char *line);

  cudaError_t qudaLaunchKernel(const void* func, dim3 gridDim, dim3 blockDim, void** args, size_t sharedMem, cudaStream_t stream);

  cudaError_t qudaEventQuery(cudaEvent_t &event);

  cudaError_t qudaEventRecord(cudaEvent_t &event, cudaStream_t stream=0);

  cudaError_t qudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags);

  cudaError_t qudaStreamSynchronize(cudaStream_t &stream);

  cudaError_t qudaEventSynchronize(cudaEvent_t &event);
  
  cudaError_t qudaDeviceSynchronize();

#if (CUDA_VERSION >= 9000)

  cudaError_t qudaFuncSetAttribute(const void* func, cudaFuncAttribute attr, int value);
#endif

  void printAPIProfile();

} // namespace quda