ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-05 06:01:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Add permuteN optimization to remove lds operation in c_shuffle (#2764)

Browse Source 
* permuteN optimization to remove lds operation in c_shuffle

* add the change log

---------

Co-authored-by: ThomasNing <thomas.ning@amd.com>
This commit is contained in:
lalala-sh
2025-09-09 13:02:48 +08:00
committed by GitHub
parent 92b07380d3
commit 75570d0fa8
5 changed files with 189 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
example/ck_tile/03_gemm/gemm_utils.hpp
View File

@@ -276,6 +276,8 @@ struct GemmConfigPreshuffleDecode : public GemmConfigBase
static constexpr ck_tile::index_t Pipeline = CK_TILE_PIPELINE_PRESHUFFLE_V2;
static constexpr bool Preshuffle = true;
static constexpr bool DoubleSmemBuffer = true;
static constexpr int N_Repeat = N_Tile / N_Warp_Tile / N_Warp;
static constexpr bool TiledMMAPermuteN = N_Repeat % 2 == 0;
};
template <typename PrecType>
@@ -298,6 +300,8 @@ struct GemmConfigPreshufflePrefill : public GemmConfigBase
static constexpr ck_tile::index_t Pipeline = CK_TILE_PIPELINE_PRESHUFFLE_V2;
static constexpr bool Preshuffle = true;
static constexpr bool DoubleSmemBuffer = true;
static constexpr int N_Repeat = N_Tile / N_Warp_Tile / N_Warp;
static constexpr bool TiledMMAPermuteN = N_Repeat % 2 == 0;
};
template <typename ADataType, typename BDataType = ADataType, typename CDataType = ADataType>