ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 13:11:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

merge flatmm pipe v0 from dteng_flatmm_opt

Browse Source
This commit is contained in:
valarLip
2025-07-23 08:44:12 +00:00
parent 6dacf833da
commit 89fa639207
5 changed files with 987 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
example/ck_tile/18_flatmm/flatmm_basic.cpp
View File

@@ -63,7 +63,7 @@ float flatmm_calc(const ck_tile::FlatmmHostArgs<>& args, const ck_tile::stream_c
using GemmPipelineProblem =
ck_tile::GemmPipelineProblem<ADataType, BDataType, AccDataType, CodegenFlatmmShape, Traits>;
using BaseGemmPipeline = ck_tile::BaseFlatmmPipelineAGmemBGmemCRegV1<GemmPipelineProblem>;
using BaseGemmPipeline = ck_tile::BaseFlatmmPipelineAGmemBGmemCRegV0<GemmPipelineProblem>;
const ck_tile::index_t k_grain = args.k_batch * FlatmmConfig::K_Tile;
const ck_tile::index_t K_split = (args.K + k_grain - 1) / k_grain * FlatmmConfig::K_Tile;
@@ -90,7 +90,7 @@ float flatmm_calc(const ck_tile::FlatmmHostArgs<>& args, const ck_tile::stream_c
tail_number_v>;
using CodegenFlatmmPipeline =
ck_tile::FlatmmPipelineAGmemBGmemCRegV1<CodegenPipelineProblem>;
ck_tile::FlatmmPipelineAGmemBGmemCRegV0<CodegenPipelineProblem>;
using GemmEpilogue = ck_tile::CShuffleEpilogue<
ck_tile::CShuffleEpilogueProblem<ADataType,