Skip to content

Re enable f8 x bf8 tests on compv3 and compv4 #3605

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
SamiAario-AMD wants to merge 4 commits into
base: develop
Choose a base branch
from
+7 −11
Open

Re enable f8 x bf8 tests on compv3 and compv4 #3605

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
a42ad99
Re-enable f8 x bf8 tests on CompV3 as they now pass
SamiAario-AMD Jan 19, 2026
4563ecb
On CompV4, fp8 x bf8 tests now pass with K_BlockSize I32
SamiAario-AMD Jan 19, 2026
c02aa36
Add a changelog entry
SamiAario-AMD Jan 19, 2026
a1e0528
Merge branch 'develop' into re-enable-f8-x-bf8-tests-on-compv3-and-co…
SamiAario-AMD Jan 20, 2026
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 CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ Documentation for Composable Kernel available at [https://rocm.docs.amd.com/proj
## Composable Kernel 1.2.0 for ROCm 7.2.0

### Added
* Added support for f8 x bf8 on CompV3, and f8 x bf8 with K_BlockSize 32 on CompV4
* Added support for bf16 data type to grouped_gemm and grouped_gemm_preshuffle.
* Added Col-Col-Row-Col layout support for aquant mode in blockscale GEMM.
* Added support for mixed precision fp8 x bf8 universal GEMM and weight preshuffle GEMM.
Expand Down
9 changes: 2 additions & 7 deletions test/ck_tile/gemm/test_gemm_pipeline_compv3.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,13 +13,8 @@ class TestCkTileGemmPipelineCompV3
static constexpr bool check_data_type()
{
using Base = TestCkTileGemmPipeline<T, TestCkTileGemmPipelineCompV3<T>>;
if constexpr(std::is_same_v<typename Base::ADataType, F8> &&
std::is_same_v<typename Base::BDataType, BF8>)
{
return false;
}
else if constexpr(std::is_same_v<typename Base::BLayout, Row> &&
std::is_same_v<typename Base::BDataType, I4>)
if constexpr(std::is_same_v<typename Base::BLayout, Row> &&
std::is_same_v<typename Base::BDataType, I4>)
{
return false;
}
Expand Down
8 changes: 4 additions & 4 deletions test/ck_tile/gemm/test_gemm_pipeline_kernel_types.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -170,7 +170,7 @@ using KernelTypesCompV4 = ::testing::Types<
std::tuple< Row, Row, Row, BF16, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, INT8, INT8, INT32, INT32, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, F8, F8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, F8, BF8, F32, F16, I256, I256, I64, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, F8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, F8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, BF8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Row, Row, BF8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
Expand All @@ -180,7 +180,7 @@ using KernelTypesCompV4 = ::testing::Types<
std::tuple< Row, Col, Row, BF16, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, INT8, INT8, INT32, INT32, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, F8, F8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, F8, BF8, F32, F16, I256, I256, I64, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, F8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, F8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, BF8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Row, Col, Row, BF8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
Expand All @@ -190,7 +190,7 @@ using KernelTypesCompV4 = ::testing::Types<
std::tuple< Col, Row, Row, BF16, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, INT8, INT8, INT32, INT32, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, F8, F8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, F8, BF8, F32, F16, I256, I256, I64, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, F8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, F8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, BF8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Row, Row, BF8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
Expand All @@ -200,7 +200,7 @@ using KernelTypesCompV4 = ::testing::Types<
std::tuple< Col, Col, Row, BF16, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, INT8, INT8, INT32, INT32, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, F8, F8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, F8, BF8, F32, F16, I256, I256, I64, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, F8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, F8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, BF8, BF8, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>,
std::tuple< Col, Col, Row, BF8, I4, F32, F16, I256, I256, I32, I32, I32, I16, Intrawave, CompV4>
Expand Down