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

Probable issue stemming through affine lowering not going through affine.apply for affine.delinearize/linearize op #952

Open
Abhishek-Varma opened this issue Dec 3, 2024 · 0 comments

Comments

@Abhishek-Varma
Copy link
Contributor

PREVIOUS STATE:
-- affine lowering used to go through affine.apply for affine.delinearize/linearize op
-- mat_truncf_affine_apply.log
-- Here is the IR snippet to focus on

CURRENT STATE:
-- affine lowering doesn't used incorporate affine.apply for affine.delinearize/linearize op's lowering.
-- mat_truncf_not_using_affiny_apply.log
-- Here is the IR snippet to focus on

For both case use the "IR snippet to focus on" as the input.mlir and run the following :-

iree-opt --allow-unregistered-dialect --canonicalize input.mlir

For "PREVIOUS STATE", you'd see no builtin.unrealized_conversion_cast remains, while that's not the case for "CURRENT STATE".

The above issue arises ONLY for "Matmul + Truncf" related CI tests as observed here in the IREE bump PR because that's where we currently deal with affine.delinearize_index due to --iree-amdaie-enable-coalescing-loops

CC: @MaheshRavishankar @krzysz00

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

No branches or pull requests

1 participant