提交 1bcecc38 authored 作者: Frederic's avatar Frederic

Added verbose command in CudaNdarray_CopyFromCudaNdarray

上级 a9e8b30c
...@@ -2837,6 +2837,9 @@ int CudaNdarray_CopyFromCudaNdarray(CudaNdarray * self, const CudaNdarray * othe ...@@ -2837,6 +2837,9 @@ int CudaNdarray_CopyFromCudaNdarray(CudaNdarray * self, const CudaNdarray * othe
} }
if (CudaNdarray_is_c_contiguous(self) && CudaNdarray_is_c_contiguous(other) && size == size_source) if (CudaNdarray_is_c_contiguous(self) && CudaNdarray_is_c_contiguous(other) && size == size_source)
{ {
if (verbose)
fprintf(stderr, "Copying contiguous vector with cublasScopy\n");
cublasScopy(size, CudaNdarray_DEV_DATA(other), 1, CudaNdarray_DEV_DATA(self), 1); cublasScopy(size, CudaNdarray_DEV_DATA(other), 1, CudaNdarray_DEV_DATA(self), 1);
CNDA_THREAD_SYNC; CNDA_THREAD_SYNC;
if (CUBLAS_STATUS_SUCCESS != cublasGetError()) if (CUBLAS_STATUS_SUCCESS != cublasGetError())
...@@ -2891,6 +2894,9 @@ int CudaNdarray_CopyFromCudaNdarray(CudaNdarray * self, const CudaNdarray * othe ...@@ -2891,6 +2894,9 @@ int CudaNdarray_CopyFromCudaNdarray(CudaNdarray * self, const CudaNdarray * othe
CudaNdarray_DEV_DATA(self), (const int *)CudaNdarray_DEV_STRIDES(self)); CudaNdarray_DEV_DATA(self), (const int *)CudaNdarray_DEV_STRIDES(self));
CNDA_THREAD_SYNC; CNDA_THREAD_SYNC;
cudaError_t err = cudaGetLastError(); cudaError_t err = cudaGetLastError();
if(verbose>1)
fprintf(stderr, "INFO k_elemwise_unary_rowmaj (n_blocks=%i, n_threads_per_block=%i)\n",
n_blocks, threads_per_block);
if( cudaSuccess != err) if( cudaSuccess != err)
{ {
//fprint_CudaNdarray(stderr, self); //fprint_CudaNdarray(stderr, self);
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论