Skip to content

Ck moe bs splitk pr #3411

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
huaiguxu wants to merge 9 commits into
base: develop
Choose a base branch
from
Open

Ck moe bs splitk pr #3411

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
cb1bea4
splitk kick-off. Compilation fail
huaiguxu Nov 26, 2025
c79879c
splitk hack pass
huaiguxu Nov 28, 2025
1e2b1f1
fix scale offset calc.
huaiguxu Dec 7, 2025
eed3c5a
Merge branch 'ck_moe_bs_splitk' into ck_moe_bs_splitk_pr
huaiguxu Dec 12, 2025
030cf6e
clang-format for a8w8_moe_blk_gemm1 splitk change
huaiguxu Dec 12, 2025
f3162c0
Merge branch 'develop' into ck_moe_bs_splitk_pr
huaiguxu Dec 15, 2025
d37c19e
Merge branch 'develop' of github.com:ROCm/composable_kernel into ck_m…
yadaish Dec 16, 2025
213f10f
fix testcase error
yadaish Dec 16, 2025
3850e1b
Merge branch 'develop' into ck_moe_bs_splitk_pr
coderfeli Dec 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions example/65_gemm_multiply_multiply/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ add_example_executable(example_moe_gemm1_xdl_fp8 moe_gemm1_xdl_fp8.cpp)
add_example_executable(example_moe_gemm2_xdl_fp8 moe_gemm2_xdl_fp8.cpp)
add_example_executable(example_moe_gemm2_xdl_fp8_blockscale moe_gemm2_xdl_fp8_blockscale.cpp)
add_example_executable(example_moe_gemm1_xdl_fp8_blockscale moe_gemm1_xdl_fp8_blockscale.cpp)
add_example_executable(example_moe_gemm1_xdl_fp8_blockscale_splitk moe_gemm1_xdl_fp8_blockscale_splitk.cpp)

list(APPEND gpu_list gfx942 gfx950 gfx1100 gfx1101 gfx1102 gfx1103 gfx1150 gfx1151 gfx1152 gfx1153 gfx1200 gfx1201 gfx11-generic gfx12-generic)
set(target 0)
Expand Down
4 changes: 2 additions & 2 deletions example/65_gemm_multiply_multiply/moe_gemm1_xdl_fp8_blockscale.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,7 +171,7 @@ using DeviceOpInstance = ck::tensor_operation::device::DeviceM
// MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector|
// PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl|
CShuffleMXDLPerWave, CShuffleNXDLPerWave, S<1, 32, 1, 8>, S<EVec, D0Vec, D1Vec, 1>,
ck::BlockGemmPipelineScheduler::Intrawave, ck::BlockGemmPipelineVersion::v1, ActOP, Nswizzle, true, MulRoutedWeight, int32_t, A0DataType>;
ck::BlockGemmPipelineScheduler::Intrawave, ck::BlockGemmPipelineVersion::v1, ActOP, Nswizzle, true, false, MulRoutedWeight, int32_t, A0DataType>;
#else
static constexpr ck::index_t MPerBlock = 64; using DeviceOpInstance = ck::tensor_operation::device::DeviceMoeGemmBlockScale<
Row, Col, DsLayout, ELayout,
Expand All @@ -185,7 +185,7 @@ static constexpr ck::index_t MPerBlock = 64; using DeviceOpInstance = ck::tensor
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 0,
S<8, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 16, 16, 0,
4, 2, S<1, 32, 1, 8>, S<2, 1, 1, 1>,
ck::BlockGemmPipelineScheduler::Intrawave, ck::BlockGemmPipelineVersion::v3, ActOP, Nswizzle, true, MulRoutedWeight, int32_t, A0DataType>;
ck::BlockGemmPipelineScheduler::Intrawave, ck::BlockGemmPipelineVersion::v3, ActOP, Nswizzle, true, false, MulRoutedWeight, int32_t, A0DataType>;
#endif
// clang-format on

Expand Down
Loading