Update fgMorphSmpOp to canonicalize commutative and compare ops at the start of POST-ORDER

[tannergooding]

No description provided.

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR adjusts CoreCLR JIT morphing to canonicalize operand ordering for commutative operators and comparisons at the start of fgMorphSmpOp post-order processing, improving consistency and enabling post-order optimizations that rely on constants being placed predictably.

Changes:

• Canonicalize commutative and comparison trees in fgMorphSmpOp post-order by swapping when the left operand is a constant (and swapping relops as needed).
• Move the ((a % pow2) ==/!= 0) => (a & (pow2-1) ==/!= 0) pattern optimization from pre-order into the GT_EQ/GT_NE post-order handling.
• Strengthen/debug-guard morphing assumptions by asserting fgMorphTree is not invoked during the value-number CSE phase.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

[tannergooding]

@EgorBo, this should be ready now. There's a few regressions where folding looks to stop happening. It needs to be investigated, but I don't think we should block on it. There's likely some other places that aren't looking for or handling the right edges.

[EgorBo]

There's a few regressions

any idea why osx-arm64 seems to be a pure regression?

[tannergooding]

any idea why osx-arm64 seems to be a pure regression?

It has more opts being done, in part because it has a higher ISA baseline.