Skip to content

Use 128-bit data loading #326

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 2, 2025
Merged

Use 128-bit data loading #326

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
4 changes: 2 additions & 2 deletions kernels/flash-attn/cutlass/flash_attn_cute.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ struct FlashAttnConfig {

// Gmem2Smem config
using GmemCopyAtom =
Copy_Atom<AutoVectorizingCopyWithAssumedAlignment<sizeof(uint128_t)>, T>;
Copy_Atom<AutoVectorizingCopyWithAssumedAlignment<sizeof(uint128_t)*8>, T>;
static constexpr int GmemValsPerLoad = sizeof(uint128_t) / sizeof(T);
static constexpr int GmemThreadsPerRow =
HeadDim / GmemValsPerLoad; // each thread reads 128 bit
Expand All @@ -56,7 +56,7 @@ struct FlashAttnConfig {
using SmemCopyAtomTransposed =
Copy_Atom<SM75_U16x8_LDSM_T, T>; // for column major load
using SmemCopyAtomO =
Copy_Atom<AutoVectorizingCopyWithAssumedAlignment<sizeof(uint128_t)>,
Copy_Atom<AutoVectorizingCopyWithAssumedAlignment<sizeof(uint128_t)*8>,
T>; // NOTE: stmatrix is only available after sm90, we use a
// vectorized copy instead

Expand Down