CommonCompareExpressionRewrite: place the negation after the op it reads - #2995
Open
wsmoses wants to merge 1 commit into
Open
CommonCompareExpressionRewrite: place the negation after the op it reads#2995wsmoses wants to merge 1 commit into
wsmoses wants to merge 1 commit into
Conversation
When the negated twin appears after the matched compare, the NotOp was created at the rewriter's default insertion point — before the compare whose result it reads — producing IR that fails dominance verification. Found via mfem QuadratureInterpolator raised kernels (#2968), where the masked strip-loop bounds emit LE/GT pairs on the same operands. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
When
CommonCompareExpressionRewritefinds the negated twin after the matched compare, it created theNotOpat the rewriter's default insertion point — before the compare whose result it reads — producing IR that fails dominance verification (operand #0 does not dominate this use).Hit by mfem's QuadratureInterpolator raised kernels under the XLA backend (#2968): masked strip-loop bounds emit
LE/GTpairs on the same operands in one block, and whether the bad arm fires depends on block order, so it surfaced as a nondeterministic compile failure.Fix: guard the insertion point and place the negation at the later compare's position. Lit test included (fails dominance verification without the fix).
🤖 Generated with Claude Code
https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD