ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-16 19:09:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

Using __builtin_amdgcn_rcpf in siLU function

Browse Source
This commit is contained in:
Qianfeng Zhang
2025-04-24 06:28:16 +00:00
parent aec19176d4
commit 7848d15d39
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/ck_tile/18_hstu_attention/hstu_attention_fwd_pipeline.hpp
View File

@@ -247,7 +247,7 @@ struct HstuAttentionFwdPipelineQRKSVS
const auto f_silu = [](CompDataType& x) {
const auto neg_one = ck_tile::type_convert<CompDataType>(-1.0f);
return x = x / (neg_one - exp(x));
x = x * __builtin_amdgcn_rcpf(neg_one - exp(x));
};
using OaccBlockTileType = decltype(gemm_1.MakeCBlockTile());