Skip to content

vulkan: Fix data race/hang in scalar/cm1 flash attention #17887

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

Open
jeffbolznv wants to merge 1 commit into
base: master
Choose a base branch
from
+6 −0
Open

vulkan: Fix data race/hang in scalar/cm1 flash attention #17887

Show file tree
Hide file tree
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
vulkan: Fix data race/hang in scalar/cm1 flash attention
  • Loading branch information
@jeffbolznv
jeffbolznv committed Dec 9, 2025
commit 67d915ac2ebd91ff44f5277b684d73284ff86a70
3 changes: 3 additions & 0 deletions ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -256,6 +256,9 @@ void main() {
barrier();
}

// prevent race on tmpsh
barrier();

// reduce across threads

[[unroll]] for (uint32_t r = 0; r < Br; ++r) {
Expand Down
3 changes: 3 additions & 0 deletions ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -302,6 +302,9 @@ void main() {
barrier();
}

// prevent race on tmpsh
barrier();

// reduce across threads

float rowmaxf[rows_per_thread], eMf[rows_per_thread], Moldf[rows_per_thread];
Expand Down
Loading