Skip to content

[Bugfix] Update accumulation handling for mma fallback in gemm_sm90.h #603

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

Merged
merged 1 commit into from
Jun 27, 2025
Merged

[Bugfix] Update accumulation handling for mma fallback in gemm_sm90.h #603

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 3 additions & 3 deletions src/tl_templates/cuda/gemm_sm90.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,7 +415,7 @@ class GemmTensorOp {
auto tCrA_view = make_tensor(tCrA.data(), remove_swizzle(tCrA.layout()));
auto tCrB_view = make_tensor(tCrB.data(), remove_swizzle(tCrB.layout()));
if constexpr (clear_accum) {
tiled_mma.accumulate_ = GMMA::ScaleOut::Zero;
clear(acc);
}
CUTE_UNROLL
for (int k = 0; k < size<2>(tCrA); ++k) {
Expand Down Expand Up @@ -448,7 +448,7 @@ class GemmTensorOp {
partition_shape_A(tiled_mma, Shape<Int<M>, Int<K>>{}));
auto tCrB_view = make_tensor(tCrB.data(), remove_swizzle(tCrB.layout()));
if constexpr (clear_accum) {
tiled_mma.accumulate_ = GMMA::ScaleOut::Zero;
clear(acc);
}
copy(tiled_copy_B, tCsB(_, _, 0), tCrB_copy_view(_, _, 0));
CUTE_UNROLL
Expand Down Expand Up @@ -483,7 +483,7 @@ class GemmTensorOp {
partition_shape_B(tiled_mma, Shape<Int<N>, Int<K>>{}));
auto tCrA_view = make_tensor(tCrA.data(), remove_swizzle(tCrA.layout()));
if constexpr (clear_accum) {
tiled_mma.accumulate_ = GMMA::ScaleOut::Zero;
clear(acc);
}
copy(tiled_copy_A, tCsA(_, _, 0), tCrA_copy_view(_, _, 0));
CUTE_UNROLL
Expand Down