[PATCH] D46086: [InstCombine] refine select-of-constants to bitwise ops

[Sanjay Patel via Phabricator via llvm-commits]

spatel added a comment.

Ping.

This patch improves the motivating cases (3-way-compare) without the regressions in https://reviews.llvm.org/D45862.

This doesn't preclude more canonicalization to select, but if we want any more select IR than this, then I think we must add folds to the DAGCombiner first.


https://reviews.llvm.org/D46086