ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Multiple-D GEMM example (#2219)

Browse Source 
* Multiple d, initial commit

* Check Ds Layout

* Readme and clang format

* Update branch & conflicts

* Multiple D - fix clang-formatter

* Rename elemetwise_op

* Fix CI

* Code review part1

* Remove printf

* Remove unnecessary comment

* Add new tests with Col layout

* Review part 2

* Added support for Multiple D GEMM

* Update comment

* Remove maybe_unused

* Clang-format

* Review part 3

* Add comment to function

* Add comment to function: another

* Take number of params for a refrence function

* Remove additional d param for 0 tensor

* Change name of function

* Fix CI fails
This commit is contained in:
Mateusz Ozga
2025-06-13 19:39:11 +02:00
committed by GitHub
parent 3a0cb27966
commit bd96ac9742
34 changed files with 2267 additions and 285 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
example/ck_tile/03_gemm/universal_gemm.cpp
View File

@@ -15,13 +15,17 @@
template <typename ADataType,
typename BDataType,
typename DsDataType,
typename AccDataType,
typename CDataType,
typename ALayout,
typename BLayout,
typename CLayout,
bool Persistent>
float gemm_calc(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
typename DsLayout,
typename ELayout,
bool Persistent,
typename CDEElementWise = ck_tile::element_wise::PassThrough>
float gemm(const ck_tile::GemmHostArgs</*NumDTensor = 0*/>& args, const ck_tile::stream_config& s)
{
using GemmShape = ck_tile::TileGemmShape<
ck_tile::sequence<GemmConfig::M_Tile, GemmConfig::N_Tile, GemmConfig::K_Tile>,
@@ -30,24 +34,26 @@ float gemm_calc(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config&
sequence<GemmConfig::M_Warp_Tile, GemmConfig::N_Warp_Tile, GemmConfig::K_Warp_Tile>,
GemmConfig::PermuteA,
GemmConfig::PermuteB>;
using TilePartitioner =
ck_tile::GemmSpatiallyLocalTilePartitioner<GemmShape,
GemmConfig::TileParitionerGroupNum,
GemmConfig::TileParitionerM01>;
using Traits = ck_tile::TileGemmTraits<GemmConfig::kPadM,
using Traits = ck_tile::TileGemmTraits<GemmConfig::kPadM,
GemmConfig::kPadN,
GemmConfig::kPadK,
ALayout,
BLayout,
CLayout>;
ELayout>;
using GemmUniversalTraits = ck_tile::TileGemmUniversalTraits<GemmConfig::kPadM,
GemmConfig::kPadN,
GemmConfig::kPadK,
GemmConfig::DoubleSmemBuffer,
ALayout,
BLayout,
CLayout,
ELayout,
GemmConfig::TransposeC,
GemmConfig::UseStructuredSparsity,
Persistent,
@@ -85,9 +91,12 @@ float gemm_calc(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config&
using GemmEpilogue = ck_tile::CShuffleEpilogue<
ck_tile::CShuffleEpilogueProblem<ADataType,
BDataType,
DsDataType,
AccDataType,
CDataType,
CLayout,
DsLayout,
ELayout,
CDEElementWise,
GemmPipelineProblem::kBlockSize,
TilePartitioner::MPerBlock,
TilePartitioner::NPerBlock,