Added support for Column Major C [Bias] in GEMM #656
+90
−50
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
kausikmaiti
force-pushed
the
column_major
branch
2 times, most recently
from
b0e4187 to
31cdc64
Compare
Signed-off-by: Maiti, Kausik <[email protected]>
kausikmaiti
force-pushed
the
column_major
branch
from
31cdc64 to
d19437b
Compare
kausikmaiti
changed the title
kausikmaiti
requested review from
Antonyvance,
amitchawla1,
anamikac-intel,
nsingh-habana,
rolandschulz and
tdeng5
petercad
reviewed
Dec 9, 2025
| std::cout << "\n\nRunning BMG GEMM with bfloat16, RowMajor Bias and bfloat16, RowMajor Output" << std::endl << std::flush; | ||
| test_bmg_gemm<bfloat16_t, cutlass::layout::RowMajor, bfloat16_t>(options, hw_info); | ||
| std::cout << "\n\nRunning BMG GEMM with bfloat16, ColumnMajor Bias and bfloat16, RowMajor Output" << std::endl << std::flush; | ||
| test_bmg_gemm<bfloat16_t, cutlass::layout::ColumnMajor, bfloat16_t>(options, hw_info); |
There was a problem hiding this comment.
I'm not sure if the 00_bmg_gemm example is the right place to test all these options -- as the very first example, I think it should probably be the simplest one. These feels like something that belongs as a test, or maybe we could have a separate 00_bmg_gemm_bias or 00_bmg_gemm_with_beta executable?
petercad
reviewed
Dec 9, 2025
|
|
||
| using ActualGmemTiledCopyC = replace_void_t<CopyOpG2R, DefaultCopyOpG2R>; | ||
| constexpr bool IsColMajorC = cutlass::gemm::detail::is_major<0, StrideC>(); | ||
| using ActualGmemTiledCopyC = replace_void_t<CopyOpG2R, std::conditional_t<IsColMajorC, CopyOpG2RTransposed, DefaultCopyOpG2R>>; |
There was a problem hiding this comment.
Suggest refactoring this a bit to make the logic/naming clearer:
using DefaultCopyOpR2GNontranspose = XE_STORE_2D<CopyBitsD, cute::gcd(8, get<0>(EpilogueTile{})), cute::gcd(512 / CopyBitsD, get<1>(EpilogueTile{}))>;
using DefaultCopyOpR2GTranspose = XE_LOAD_2D_TRANSPOSE<CopyBitsCTranspose, cute::gcd(512 / CopyBitsC, get<1>(EpilogueTile{})), cute::gcd(8 / Sub32BitFactor, get<0>(EpilogueTile{}))>;
using DefaultCopyOpR2G = conditional_t<IsColMajorC, DefaultCopyOpR2GTranspose, DefaultCopyOpR2GNontranspose>;
using ActualGmemTiledCopyC = replace_void_t<CopyOpG2R, DefaultCopyOpG2R>;
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Added support for Column Major C [Bias] in GEMM.
Following changes have been made.
To do: