scemama/qp_plugins_scemama
1
0
Fork 0
mirror of https://gitlab.com/scemama/qp_plugins_scemama.git synced 2024-06-02 11:25:23 +02:00
Code Issues Releases Wiki Activity

Added gpu_malloc in gpu_dgemm

Browse Source
This commit is contained in:
Anthony Scemama 2023-12-06 21:40:20 +01:00
parent 29903111f4
commit b557b57a3b
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
devel/ccsd_gpu/gpu_dgemm.c
View File

@ -27,13 +27,13 @@ void gpu_dgemm(char transa, char transb, int m, int n, int k, double alpha,
cublasOperation_t ta, tb; cublasOperation_t ta, tb;
if (transa == 'N') { if (transa == 'N') {
cudaStat = cudaMalloc((void**)&d_A, (size_t) lda*k*sizeof(double)); cudaStat = gpu_malloc((void**)&d_A, (size_t) lda*k*sizeof(double));
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
cudaStat = cublasSetMatrix(m, k, sizeof(double), A, lda, d_A, lda); cudaStat = cublasSetMatrix(m, k, sizeof(double), A, lda, d_A, lda);
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
ta = CUBLAS_OP_N; ta = CUBLAS_OP_N;
} else { } else {
cudaStat = cudaMalloc((void**)&d_A, (size_t) lda*m*sizeof(double)); cudaStat = gpu_malloc((void**)&d_A, (size_t) lda*m*sizeof(double));
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
cudaStat = cublasSetMatrix(k, m, sizeof(double), A, lda, d_A, lda); cudaStat = cublasSetMatrix(k, m, sizeof(double), A, lda, d_A, lda);
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
@ -41,20 +41,20 @@ void gpu_dgemm(char transa, char transb, int m, int n, int k, double alpha,
} }
if (transb == 'N') { if (transb == 'N') {
cudaStat = cudaMalloc((void**)&d_B, (size_t) ldb*n*sizeof(double)); cudaStat = gpu_malloc((void**)&d_B, (size_t) ldb*n*sizeof(double));
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
cudaStat = cublasSetMatrix(k, n, sizeof(double), B, ldb, d_B, ldb); cudaStat = cublasSetMatrix(k, n, sizeof(double), B, ldb, d_B, ldb);
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
tb = CUBLAS_OP_N; tb = CUBLAS_OP_N;
} else { } else {
cudaStat = cudaMalloc((void**)&d_B, (size_t) ldb*k*sizeof(double)); cudaStat = gpu_malloc((void**)&d_B, (size_t) ldb*k*sizeof(double));
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
cudaStat = cublasSetMatrix(n, k, sizeof(double), B, ldb, d_B, ldb); cudaStat = cublasSetMatrix(n, k, sizeof(double), B, ldb, d_B, ldb);
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
tb = CUBLAS_OP_T; tb = CUBLAS_OP_T;
} }
cudaStat = cudaMalloc((void**)&d_C, (size_t) ldc*n*sizeof(double)); cudaStat = gpu_malloc((void**)&d_C, (size_t) ldc*n*sizeof(double));
assert(cudaStat == cudaSuccess); assert(cudaStat == cudaSuccess);
if (beta != 0.) { if (beta != 0.) {
cudaStat = cublasSetMatrix(m, n, sizeof(double), C, ldc, d_C, ldc); cudaStat = cublasSetMatrix(m, n, sizeof(double), C, ldc, d_C, ldc);