diff --git a/devel/ccsd_gpu/gpu_dgemm.c b/devel/ccsd_gpu/gpu_dgemm.c index 9ade048..d824da3 100644 --- a/devel/ccsd_gpu/gpu_dgemm.c +++ b/devel/ccsd_gpu/gpu_dgemm.c @@ -27,13 +27,13 @@ void gpu_dgemm(char transa, char transb, int m, int n, int k, double alpha, cublasOperation_t ta, tb; 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); cudaStat = cublasSetMatrix(m, k, sizeof(double), A, lda, d_A, lda); assert(cudaStat == cudaSuccess); ta = CUBLAS_OP_N; } 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); cudaStat = cublasSetMatrix(k, m, sizeof(double), A, lda, d_A, lda); 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') { - 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); cudaStat = cublasSetMatrix(k, n, sizeof(double), B, ldb, d_B, ldb); assert(cudaStat == cudaSuccess); tb = CUBLAS_OP_N; } 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); cudaStat = cublasSetMatrix(n, k, sizeof(double), B, ldb, d_B, ldb); assert(cudaStat == cudaSuccess); 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); if (beta != 0.) { cudaStat = cublasSetMatrix(m, n, sizeof(double), C, ldc, d_C, ldc);