ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-21 05:19:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix tail handler bug

Browse Source
This commit is contained in:
lalala-sh
2025-07-17 08:40:35 +00:00
parent fb76450e63
commit 9aa3396a79
4 changed files with 19 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/ck_tile/18_flatmm/flatmm_basic.cpp
View File

@@ -80,7 +80,7 @@ float flatmm_calc(const ck_tile::FlatmmHostArgs<>& args, const ck_tile::stream_c
constexpr auto scheduler = FlatmmConfig::Scheduler;
constexpr auto memory_operation = memory_operation_.value;
using CodegenPipelineProblem = ck_tile::UniversalGemmPipelineProblem<ADataType,
using CodegenPipelineProblem = ck_tile::FlatmmPipelineProblem<ADataType,
BDataType,
AccDataType,
CodegenFlatmmShape,

3
example/ck_tile/18_flatmm/flatmm_basic.hpp
View File

@@ -83,7 +83,10 @@ struct FlatmmConfig16
template <typename DataType>
struct FlatmmConfig16_950 : public FlatmmConfig16<DataType>
{
static constexpr ck_tile::index_t N_Tile = 256;
static constexpr ck_tile::index_t K_Tile = 256 / sizeof(DataType);
static constexpr ck_tile::index_t K_Warp_Tile = sizeof(DataType) == 2 ? 32 : 128;
static constexpr int kBlockPerCu = 1;
};
template <typename ADataType>