Skip to content

Conversation

@borontion
Copy link
Contributor

Given we already encode the kDim in instrShape, we don't need to select again when lowering mfma, just directly get the instruction, when matching all parameters.

Copy link
Collaborator

@antiagainst antiagainst left a comment

Choose a reason for hiding this comment

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

Thanks for cleaning this up!

@antiagainst antiagainst changed the title [AMD] Replace wmma select [AMD] Replace mfma select in LLVM conversion Sep 30, 2025
@antiagainst antiagainst marked this pull request as ready for review September 30, 2025 05:28
@antiagainst antiagainst enabled auto-merge (squash) September 30, 2025 05:30
auto-merge was automatically disabled September 30, 2025 05:46

Head branch was pushed to by a user without write access

@borontion borontion requested a review from ptillet as a code owner September 30, 2025 05:46
@antiagainst antiagainst merged commit 8b8c607 into triton-lang:main Sep 30, 2025
9 checks passed
@borontion borontion deleted the amd-mfma-kdim branch September 30, 2025 23:21
yangshuxin pushed a commit to yangshuxin/triton that referenced this pull request Oct 3, 2025
Given we already encode the kDim in `instrShape`, we don't need to
select again when lowering mfma, just directly get the instruction, when
matching all parameters.
Jokeren pushed a commit that referenced this pull request Oct 10, 2025
Given we already encode the kDim in `instrShape`, we don't need to
select again when lowering mfma, just directly get the instruction, when
matching all parameters.
ita9naiwa pushed a commit to ita9naiwa/triton that referenced this pull request Oct 12, 2025
Given we already encode the kDim in `instrShape`, we don't need to
select again when lowering mfma, just directly get the instruction, when
matching all parameters.
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