Skip to content

[CuTe,Sm110] Fix sm110 2cta dQ postprocess #2491

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

Closed
gpzlx1 wants to merge 1 commit into from
+2 −2
Closed

[CuTe,Sm110] Fix sm110 2cta dQ postprocess #2491

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 flash_attn/cute/flash_bwd_postprocess.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@ def __init__(
self.num_threads = num_threads
self.AtomLayoutMdQ = AtomLayoutMdQ
self.dQ_swapAB = dQ_swapAB
self.use_2cta_instrs = use_2cta_instrs and arch // 10 == 10 and head_dim != 64
self.use_2cta_instrs = use_2cta_instrs and arch // 10 in [10, 11] and head_dim != 64
self.cluster_size = cluster_size

@staticmethod
Expand Down Expand Up @@ -373,7 +373,7 @@ def kernel(
seqlen_q = seqlen.seqlen_q
seqlen_q_rounded = cute.round_up(seqlen_q, self.tile_m)

if const_expr(self.arch // 10 == 10 and self.use_2cta_instrs):
if const_expr(self.arch // 10 in [10, 11] and self.use_2cta_instrs):
# 2-CTA: remap dQaccum layout into TMEM view before writing sdQ
num_reduce_threads = self.num_threads
thr_mma_dsk = tiled_mma.get_slice(tidx)
Expand Down