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

Preshuffle Decode Prefill config fix (#2693)

Browse Source 
* feat(gemm_wp): add two new configs for wp

* delete the unnecessary files

* fix the config error

* update the config

---------

Co-authored-by: AviralGoelAMD <aviral.goel@amd.com>
This commit is contained in:
Thomas Ning
2025-08-15 15:49:07 -07:00
committed by GitHub
parent c06e8b4a66
commit 5ada85ec04
1 changed files with 18 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
example/ck_tile/03_gemm/gemm_utils.hpp Normal file → Executable file
View File

@@ -35,6 +35,22 @@ constexpr ck_tile::index_t get_k_warp_tile()
#endif
}
template <typename PrecType, ck_tile::index_t M_Warp_Tile>
constexpr ck_tile::index_t get_k_warp_tile_flatmm()
{
#if defined(CK_GFX950_SUPPORT)
if constexpr(M_Warp_Tile == 32)
return sizeof(PrecType) == 2 ? 16 : 64;
else
return sizeof(PrecType) == 2 ? 32 : 128;
#else
if constexpr(M_Warp_Tile == 32)
return sizeof(PrecType) == 2 ? 16 : 32;
else
return sizeof(PrecType) == 2 ? 32 : 64;
#endif
}
struct GemmConfigBase
{
static constexpr bool kPadM = false;
@@ -229,7 +245,7 @@ struct GemmConfigPreshuffleDecode : public GemmConfigBase
static constexpr ck_tile::index_t M_Warp_Tile = 16;
static constexpr ck_tile::index_t N_Warp_Tile = 16;
static constexpr ck_tile::index_t K_Warp_Tile = get_k_warp_tile<PrecType, M_Warp_Tile>();
static constexpr ck_tile::index_t K_Warp_Tile = get_k_warp_tile_flatmm<PrecType, M_Warp_Tile>();
static constexpr int kBlockPerCu = 1;
static constexpr auto Scheduler = ck_tile::GemmPipelineScheduler::Default;
@@ -251,7 +267,7 @@ struct GemmConfigPreshufflePrefill : public GemmConfigBase
static constexpr ck_tile::index_t M_Warp_Tile = 16;
static constexpr ck_tile::index_t N_Warp_Tile = 16;
static constexpr ck_tile::index_t K_Warp_Tile = get_k_warp_tile<PrecType, M_Warp_Tile>();
static constexpr ck_tile::index_t K_Warp_Tile = get_k_warp_tile_flatmm<PrecType, M_Warp_Tile>();
static constexpr int kBlockPerCu = 2;
static constexpr auto Scheduler = ck_tile::GemmPipelineScheduler::Default;