Skip to content

UPSTREAM PR #18785: CUDA: Factor out and re-use two_stage_warp_reduce function #897

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
loci-dev wants to merge 2 commits into
base: main
Choose a base branch
from

Use `two_stage_warp_reduce` also in softmax kernel, move smem out of it

1ebe58d
Select commit
Loading
Failed to load commit list.
Open

UPSTREAM PR #18785: CUDA: Factor out and re-use two_stage_warp_reduce function #897

Use `two_stage_warp_reduce` also in softmax kernel, move smem out of it
1ebe58d
Select commit
Loading
Failed to load commit list.
LOCI Review / Performance Review #897 succeeded Jan 12, 2026

Performance unchanged

0 binaries improved · 0 binaries unchanged · 0 binaries stable ~ within threshold · 0 binaries degraded ~ beyond threshold

Binary Δ % Response Δ % Throughput Performance (based on response time)

Performance threshold: 30%
Default configuration used.
Note: Performance status is evaluated only from Δ% Response. Throughput is displayed for reference.

Explore the complete analysis inside the Version Insights.
Open the Pull Request linked to this check-run.

View more details on LOCI Review