diff options
| author | aamir <[email protected]> | 2018-08-13 14:57:41 -0700 |
|---|---|---|
| committer | aamir <[email protected]> | 2018-08-13 14:57:41 -0700 |
| commit | 44f0114ad2c208f69c0c1baa980a5b3bda37e16b (patch) | |
| tree | 268edd06a4dbc9ca2977e73274f4da983b3b3153 /cuda-kernels/v4p_genericMatrixMultiply.cu | |
| parent | 35695a3fb6dee0ad94baaa97dd8d7a9ad9d8156d (diff) | |
generalized v8 and v16 mode kernel completed
Diffstat (limited to 'cuda-kernels/v4p_genericMatrixMultiply.cu')
| -rw-r--r-- | cuda-kernels/v4p_genericMatrixMultiply.cu | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/cuda-kernels/v4p_genericMatrixMultiply.cu b/cuda-kernels/v4p_genericMatrixMultiply.cu index a08903b..1b56eb2 100644 --- a/cuda-kernels/v4p_genericMatrixMultiply.cu +++ b/cuda-kernels/v4p_genericMatrixMultiply.cu @@ -20,9 +20,9 @@ void curandErrCheck_(curandStatus_t stat, const char *file, int line) { using namespace nvcuda; // Must be multiples of 16 for wmma code to work -#define MATRIX_M (32) -#define MATRIX_N (32) -#define MATRIX_K (32) +#define MATRIX_M (1024) +#define MATRIX_N (1024) +#define MATRIX_K (1024) // The only dimensions currently supported by WMMA |