ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-15 10:37:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

Set the block_per_cu to 3 for hdim-128

Browse Source
This commit is contained in:
Qianfeng Zhang
2025-05-18 15:58:02 +00:00
parent ff3415d97d
commit e4e70f8b0a
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp
View File

@@ -84,9 +84,9 @@ struct HstuAttentionFwdPipelineQRKSVS
else if constexpr(kQKHeaddim == 96 || kQKHeaddim == 128)
{
if constexpr(kHasBias)
return 1;
else
return 2;
else
return 3;
}
else if constexpr(kQKHeaddim == 256)
{