Skip to content

[EP] Add benchmarking option for FP8 dispatch.#743

Merged
laochonlam merged 2 commits intomainfrom
feat/ll-dispatch-fp8-toggle
Feb 22, 2026
Merged

[EP] Add benchmarking option for FP8 dispatch.#743
laochonlam merged 2 commits intomainfrom
feat/ll-dispatch-fp8-toggle

Conversation

@laochonlam
Copy link
Copy Markdown
Collaborator

@laochonlam laochonlam commented Feb 22, 2026

Enable benchmarking with either FP8 dispatch or direct BF16 dispatch through a runtime flag for apples-to-apples performance comparison.

Type of Change

  • Bug fix
  • New feature
  • Documentation update

How Has This Been Tested?

Include any tests here.

  • Unit tests
  • Integration tests
  • Manual testing

Checklist

  • My code follows the style guidelines, e.g. format.sh.
  • I have run build_and_install.sh to verify compilation.
  • I have removed redundant variables and comments.
  • I have updated the documentation.
  • I have added tests.

Enable benchmarking with either FP8 dispatch or direct BF16 dispatch through a runtime flag for apples-to-apples performance comparison.
@laochonlam laochonlam requested a review from MaoZiming February 22, 2026 00:13
@laochonlam laochonlam changed the title Add benchmarking option for FP8 dispatch. [EP] Add benchmarking option for FP8 dispatch. Feb 22, 2026
Copy link
Copy Markdown
Member

@MaoZiming MaoZiming left a comment

Choose a reason for hiding this comment

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

Feel free to merge after you have run format.sh

Keep behavior unchanged while making the conditional tuple iteration clearer and Black-compliant.

Co-authored-by: Cursor <cursoragent@cursor.com>
@laochonlam laochonlam merged commit 579b98c into main Feb 22, 2026
3 checks passed
@laochonlam laochonlam deleted the feat/ll-dispatch-fp8-toggle branch February 22, 2026 01:07
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants