Skip to content

[DEBUG] Disable loop versioning to check compile time #4043

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.

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

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

anmyachev
Copy link
Contributor

@anmyachev anmyachev commented Apr 28, 2025

  • test_scaled_dot does not include a loop, so versioning does not affect its execution time
  • test_mxfp includes a loop, I ran one test combination test_mxfp[0-4-3-128-256-128-2-4-32] and got the following numbers: 22.06s (turned off versioning) vs 30.33s (turned on versioning). This time includes compilation (clean cache) and the launch itself.

сс @whitneywhtsang

@whitneywhtsang
Copy link
Contributor

@etiotto Do you think we should limit the RemoveMasks pass to only LoadOp or StoreOp with BlockIO attribute?

@etiotto
Copy link
Contributor

etiotto commented Apr 29, 2025

@etiotto Do you think we should limit the RemoveMasks pass to only LoadOp or StoreOp with BlockIO attribute?

The RemoveMask pass is a TT IR pass so at the point it is run the loads aren't marked with that attribute.
Perhaps the axisInfo analysis can be used at that point in the pass pipeline. I'd suggest opening an issue. This isn't urgent at this point but is good to have an issue so we can try to improve it later.

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.

3 participants