Skip to content
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

AIRCC: Remove experimental passes from aircc.py #908

Conversation

erwei-xilinx
Copy link
Collaborator

Those passes are now default passes for npu.

Also added a number of flags to control whether some passes get called.

Copy link
Collaborator

@fifield fifield left a comment

Choose a reason for hiding this comment

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

I'm happy to see that the pass pipeline has graduated from "experimental".

Comment on lines 48 to +53
verbose=False,
experimental_passes=False,
omit_while_true_loop=False,
omit_pingpong=False,
lower_linalg_to_func=False,
air_loop_fusion=False,
runtime_loop_tiling_sizes: list[int] = [4, 4],
Copy link
Collaborator

Choose a reason for hiding this comment

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

We got rid of one option but grew some new ones :/

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Yeah... But the good thing is that they now more clearly specify what "experimental" features were being used. The lower_linalg_to_func is likely to be necessary to say if we want micro-kernel or direct code-generated (scalar, for now) code.

I wish the air-loop-fusion and runtim-loop-tiling-sizes can go eventually...

@erwei-xilinx erwei-xilinx merged commit 54cf776 into Xilinx:main Feb 22, 2025
11 checks passed
@erwei-xilinx erwei-xilinx deleted the move_experimental_passes_to_default_npu_passes branch February 22, 2025 05:17
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