ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-18 03:49:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

refactor

Browse Source 
[ROCm/composable_kernel commit: 4543d17a71]
This commit is contained in:
Chao Liu
2019-02-19 22:07:15 -06:00
parent 23b7f3ef33
commit 7080421844
3 changed files with 6 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/include/gridwise_implicit_gemm_convolution_1_chwn_csrk_khwn.hip.hpp
View File

@@ -94,8 +94,8 @@ gridwise_implicit_gemm_convolution_1_chwn_csrk_khwn(const Float* const __restric
// tensor view of blockwise input and weight in LDS
// be careful of alignment
constexpr auto in_chwn_block_desc =
make_ConstantTensorDescriptor(Sequence<CPerBlock, HiPerBlock, WiPerBlock, NPerBlock>{});
constexpr auto in_chwn_block_desc = make_ConstantTensorDescriptor_aligned(
Sequence<CPerBlock, HiPerBlock, WiPerBlock, NPerBlock>{}, Number<InBlockCopyDataPerRead>{});
constexpr auto wei_ek_block_desc = make_ConstantTensorDescriptor_aligned(
Sequence<CPerBlock * S * R, KPerBlock>{}, Number<WeiBlockCopyDataPerRead>{});
@@ -164,7 +164,9 @@ gridwise_implicit_gemm_convolution_1_chwn_csrk_khwn(const Float* const __restric
HoPerThread>{};
// LDS: be careful of alignment
constexpr unsigned in_block_size = in_chwn_block_desc.GetElementSpace();
constexpr unsigned in_block_size =
in_chwn_block_desc.GetElementSpace(Number<InBlockCopyDataPerRead>{});
constexpr unsigned wei_block_size =
wei_csrk_block_desc.GetElementSpace(Number<WeiBlockCopyDataPerRead>{});