Skip to content

Conversation

@cagrikymk
Copy link
Contributor

This PR adds more optimizations for the PE case of the MHA, for more details about the computation: #1184

Optimizations:

  • New config.
  • Preloading v
  • Reordering pe matmul
  • Simplifying scaling ops (for sm_scale and log_2(e) related scaling)
  • Masking related changes

@cagrikymk cagrikymk changed the title [WIP][TRITON] Cagri/mha pa optimizations [WIP][TRITON] MHA PA optimizations Oct 23, 2025
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.

1 participant