Skip to content

Clamp SM90 tiles to GMMA-friendly sizes #15

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

Jump to bottom
Merged
ivandobskygithub merged 2 commits into from
Nov 25, 2025
Merged

Clamp SM90 tiles to GMMA-friendly sizes #15

ivandobskygithub merged 2 commits into from
Nov 25, 2025

Conversation

@ivandobskygithub
Copy link
Owner

@ivandobskygithub ivandobskygithub commented Nov 25, 2025

Summary

  • align SM90 shared-memory clamping to 16-wide tiles that satisfy GMMA copy requirements and shrink BlockM when needed
  • soften the shared-memory estimate for very large head/value dimensions to match the single-stage SM120 pipeline
  • update the shared-memory budget test harness to mirror the buffering-aware estimate

Testing

  • pytest tests/hopper/test_tile_size_shared_memory.py -q

Codex Task

@ivandobskygithub ivandobskygithub merged commit 31e3680 into main Nov 25, 2025
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

codex

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@ivandobskygithub