8000 vulkan: use scalar FA rather than coopmat2 when N==1 by jeffbolznv · Pull Request #13554 · ggml-org/llama.cpp · GitHub
[go: up one dir, main page]
Skip to content

vulkan: use scalar FA rather than coopmat2 when N==1 #13554

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
May 17, 2025
Merged

vulkan: use scalar FA rather than coopmat2 when N==1 #13554

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
7 changes: 7 additions & 0 deletions ggml/src/ggml-vulkan/ggml-vulkan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5872,10 +5872,17 @@ static void ggml_vk_flash_attn(ggml_backend_vk_context * ctx, vk_context& subctx
vk_pipeline *pipelines;
bool small_rows = N <= get_fa_num_small_rows(path);

// coopmat1 does not actually support "small rows" (it needs 16 rows).
// So use scalar instead.
if (small_rows && path == FA_COOPMAT1) {
path = FA_SCALAR;
}

// scalar is faster than coopmat2 when N==1
if (N == 1 && path == FA_COOPMAT2) {
path = FA_SCALAR;
}

bool f32acc = path == FA_SCALAR || dst->op_params[3] == GGML_PREC_F32;

switch (path) {
Expand Down
Loading
0