ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-17 19:40:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Xdlops refactor fix (#22)

Browse Source 
* added constexpr ahead of adptor; clean unused driver; rename M/NPerWave to M/NPerXDL

* fixed bwd

* fixed comment

[ROCm/composable_kernel commit: 9d3f634a3c]
This commit is contained in:
zjing14
2021-08-23 11:22:10 -05:00
committed by GitHub
parent eda5a8852c
commit 14e4d7ba7d
11 changed files with 111 additions and 383 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
host/solver/include/conv_tunable_fwd_v4r4_xdlops_nchw_kcyx_nkhw.hpp
View File

@@ -9,8 +9,8 @@ struct tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw
int NPerBlock;
int KPerBlock;
int MPerWave;
int NPerWave;
int MPerXDL;
int NPerXDL;
int K1;
int MRepeat;
@@ -45,8 +45,8 @@ static tunable_dyn_conv_fwd_v4r4_xdlops_nchw_kcyx_nkhw
128, // MPerBlock,
128, // NPerBlock,
4, // KPerBlock,
32, // MPerWave,
32, // NPerWave,
32, // MPerXDL,
32, // NPerXDL,
4, // K1,
2, // MRepeat,
2, // NRepeat,