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

fix flatmm kernel for bigger size for fp16 datatype (#2302)

Browse Source
This commit is contained in:
Khushbu Agarwal
2025-06-10 11:13:40 -07:00
committed by GitHub
parent aed0f5880c
commit bd270fe4bc
6 changed files with 91 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
example/ck_tile/18_flatmm/flatmm_basic.hpp
View File

@@ -109,6 +109,43 @@ struct is_8bit_type
{
};
template <typename ADataType>
struct FlatmmConfig
{
#if defined(USING_MFMA_16x16x32)
static constexpr ck_tile::index_t M_Tile = 128;
static constexpr ck_tile::index_t N_Tile = 128;
static constexpr ck_tile::index_t K_Tile = 128;
static constexpr ck_tile::index_t M_Warp = 1;
static constexpr ck_tile::index_t N_Warp = 4;
static constexpr ck_tile::index_t K_Warp = 1;
static constexpr ck_tile::index_t M_Warp_Tile = is_8bit_type<ADataType>::value ? 16 : 32;
static constexpr ck_tile::index_t N_Warp_Tile = is_8bit_type<ADataType>::value ? 16 : 32;
static constexpr ck_tile::index_t K_Warp_Tile = is_8bit_type<ADataType>::value ? 64 : 16;
#elif defined(USING_MFMA_32x32x16)
static constexpr ck_tile::index_t M_Tile = 128;
static constexpr ck_tile::index_t N_Tile = 256;
static constexpr ck_tile::index_t K_Tile = 128;
static constexpr ck_tile::index_t M_Warp = 1;
static constexpr ck_tile::index_t N_Warp = 8;
static constexpr ck_tile::index_t K_Warp = 1;
static constexpr ck_tile::index_t M_Warp_Tile = is_8bit_type<ADataType>::value ? 32 : 32;
static constexpr ck_tile::index_t N_Warp_Tile = is_8bit_type<ADataType>::value ? 32 : 32;
static constexpr ck_tile::index_t K_Warp_Tile = is_8bit_type<ADataType>::value ? 32 : 16;
#endif
// The kPadM, kPadN, kPadK & kBlockPerCu should also come from the Codegen part.
static constexpr bool kPadM = false;
static constexpr bool kPadN = false;
static constexpr bool kPadK = false;
static constexpr int kBlockPerCu = 2;
};
auto create_args(int argc, char* argv[])
{
ck_tile::ArgParser arg_parser;