Skip to content

[Codegen][GPU] Add kernel config for ArgCompareOp with VectorDistribute pipeline #24003

Open
bangtianliu wants to merge 4 commits intoiree-org:mainfrom
bangtianliu:argcompare-kernel-config-v2
Open

[Codegen][GPU] Add kernel config for ArgCompareOp with VectorDistribute pipeline #24003
bangtianliu wants to merge 4 commits intoiree-org:mainfrom
bangtianliu:argcompare-kernel-config-v2

Conversation

@bangtianliu
Copy link
Copy Markdown
Contributor

This PR adds kernel config for iree_linalg_ext.arg_compare operations with the VectorDistribute pipeline.

Note: Currently rebase on #23793, so please only review last commit.

Issue: #23005
Assisted-by: Claude Code

Signed-off-by: Bangtian Liu <liubangtian@gmail.com>
Signed-off-by: Bangtian Liu <liubangtian@gmail.com>
  - Extract shared helpers and to deduplicate shape computation logic between DistributeMultiReduction and DistributeArgCompare
  - Simplify broadcastShape computation by deriving from distributed input shape instead of manual construction
  - Reuse existing elemTy/indexElemTy variables instead of repeated getElementType() calls
  - Simplify resultValue initialization by setting common case first
  - Fix inaccurate comment about yielded value check

Signed-off-by: Bangtian Liu <liubangtian@gmail.com>
@bangtianliu bangtianliu marked this pull request as draft April 2, 2026 23:01
@bangtianliu bangtianliu force-pushed the argcompare-kernel-config-v2 branch 3 times, most recently from a1125a0 to f82a041 Compare April 3, 2026 17:13
…te pipeline

Signed-off-by: Bangtian Liu <liubangtian@gmail.com>
@bangtianliu bangtianliu marked this pull request as ready for review April 3, 2026 17:18
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