Add gemm universal bf16 instances (#1484)

* revert ckprofiler change

* temp save

* Add test and test pass

* test pass

* Fix bug inside rotating buffer when tensor is not packed

* bug fix

* clang format

---------

Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>
This commit is contained in:
Haocong WANG
2024-09-05 11:58:54 +08:00
committed by GitHub
parent 52410b49c7
commit 5b10dae6a4
35 changed files with 1048 additions and 75 deletions

View File

@@ -171,6 +171,14 @@ int profile_gemm_universal(int argc, char* argv[])
{
return profile(BF16{}, BF16{}, BF16{}, F32{}, BF16{}, Row{}, Col{}, Row{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::KM_NK_MN)
{
return profile(BF16{}, BF16{}, BF16{}, F32{}, BF16{}, Col{}, Col{}, Row{});
}
else if(data_type == GemmDataType::BF16_BF16_BF16 && layout == GemmMatrixLayout::KM_KN_MN)
{
return profile(BF16{}, BF16{}, BF16{}, F32{}, BF16{}, Col{}, Row{}, Row{});
}
else if(data_type == GemmDataType::F8_F8_BF16 && layout == GemmMatrixLayout::MK_KN_MN)
{
return profile(F8{}, F8{}, F8{}, F32{}, BF16{}, Row{}, Row{}, Row{});