ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 01:10:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

enable 128x128 block gemm

Browse Source
This commit is contained in:
Chao Liu
2019-04-03 19:04:17 -05:00
parent 6a3f3f951d
commit 05d7a0875c
5 changed files with 112 additions and 125 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -190,7 +190,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 256;
#elif 1
#elif 0
// 1x1, 14x14, Vega 20, disable lds_double_buffer, enable register double buffer
constexpr index_t BPerBlock = 64;
constexpr index_t KPerBlock = 128;
@@ -221,8 +221,8 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t BlockSize = 128;
#elif 1
// 1x1, 14x14, Vega 20, hack CPerBlock = 1
constexpr index_t BPerBlock = 64;
// 1x1, 14x14, Vega 20, try
constexpr index_t BPerBlock = 128;
constexpr index_t KPerBlock = 128;
constexpr index_t CPerBlock = 8;