ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 09:16:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

tune implicit_gemm

Browse Source
This commit is contained in:
Chao Liu
2019-01-17 00:01:21 -06:00
parent 216e3da609
commit 3bd51021ab
4 changed files with 31 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/include/gridwise_implicit_gemm_convolution_nchw_kcsr.cuh
View File

@@ -185,7 +185,7 @@ __global__ void gridwise_implicit_gemm_convolution_nchw_kcsr(InGlobalDesc,
in_nchw_block_desc.GetLengths());
#endif
#if 0
#if 1
// weight: global mem to LDS,
// convert [K,C,S,R] to [S,R,C,K]
blockwise_4d_tensor_copy_reorder_by_get_dst_from_src<BlockSize>(
@@ -238,7 +238,7 @@ __global__ void gridwise_implicit_gemm_convolution_nchw_kcsr(InGlobalDesc,
const unsigned k_thread_data_begin = matrix_c_index.row_begin;
const unsigned wo_thread_data_begin = matrix_c_index.col_begin / NPerThread;
#if 0
#if 1
// output: register to global mem,
// convert out_thread[Ho,K,Wo,N] to out_global[N,K,Ho,Wo]
constexpr auto reorder_nkhw_from_hkwn = Sequence<3, 1, 0, 2>{};