Skip to content

Add GQA group_size 5, 6, 7 to DISPATCH_GQA_GROUP_SIZE #2986

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
arbi-dev wants to merge 1 commit into
base: main
Choose a base branch
from
+9 −0
Open

Add GQA group_size 5, 6, 7 to DISPATCH_GQA_GROUP_SIZE #2986

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
9 changes: 9 additions & 0 deletions include/flashinfer/utils.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,6 +147,15 @@
} else if (group_size == 4) { \
constexpr size_t GROUP_SIZE = 4; \
__VA_ARGS__ \
} else if (group_size == 5) { \
constexpr size_t GROUP_SIZE = 5; \
__VA_ARGS__ \
} else if (group_size == 6) { \
constexpr size_t GROUP_SIZE = 6; \
__VA_ARGS__ \
} else if (group_size == 7) { \
constexpr size_t GROUP_SIZE = 7; \
__VA_ARGS__ \
Comment on lines +150 to +158
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Apr 5, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

With the addition of more group sizes, the if-else if chain in DISPATCH_GQA_GROUP_SIZE is becoming increasingly long. Consider refactoring the macro to use a switch statement. This would ensure the group_size expression is evaluated only once and would improve consistency with other dispatch macros in this file (such as DISPATCH_CTA_TILE_Q and DISPATCH_HEAD_DIM) that already use switch for exact value matching.

} else if (group_size == 8) { \
constexpr size_t GROUP_SIZE = 8; \
__VA_ARGS__ \
Expand Down
Loading