diff options
| author | aamir <[email protected]> | 2018-06-01 19:26:04 -0700 |
|---|---|---|
| committer | aamir <[email protected]> | 2018-06-01 19:26:04 -0700 |
| commit | 57c32a360779c0b9899e692a0d60e7dbed6bb984 (patch) | |
| tree | 2b1509ddad3f2817e2990fd2e53cb7c417207d3b /cuda-kernels | |
| parent | e177edcdefca06d7d3a7bc7be5f6a759f69b909e (diff) | |
added mma & mma_st
Diffstat (limited to 'cuda-kernels')
| -rw-r--r-- | cuda-kernels/tensor_core.cu | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/cuda-kernels/tensor_core.cu b/cuda-kernels/tensor_core.cu index 483a42b..d5c1b40 100644 --- a/cuda-kernels/tensor_core.cu +++ b/cuda-kernels/tensor_core.cu @@ -156,14 +156,14 @@ int main(int argc, char* argv[]) { // printf("a_fp32\n"); for(int m=0;m<MATRIX_M;m++){ for(int n=0;n<MATRIX_K;n++){ - a_host_wmma[m*MATRIX_K+n]=m*MATRIX_K+n; + a_host_wmma[m*MATRIX_K+n]=(m*MATRIX_K+n)%10; } //printf(";\n"); } // printf("b_fp32\n"); for(int m=0;m<MATRIX_K;m++){ for(int n=0;n<MATRIX_N;n++){ - b_host_wmma[m*MATRIX_N+n]=m*MATRIX_N+n; + b_host_wmma[m*MATRIX_N+n]=(m*MATRIX_N+n)%10; // printf("%f ",b_host_wmma[m*MATRIX_N+n]); } // printf(";\n"); @@ -177,7 +177,7 @@ int main(int argc, char* argv[]) { for(int m=0;m<MATRIX_M;m++){ for(int n=0;n<MATRIX_N;n++){ - c_init_host_wmma[m*MATRIX_N+n]=0; + c_init_host_wmma[m*MATRIX_N+n]=(m*MATRIX_M+n)%10; } } cudaErrCheck(cudaMemcpy(c, c_init_host_wmma, MATRIX_M * MATRIX_N * sizeof(float), cudaMemcpyHostToDevice)); |