[Mlir-commits] [mlir] [mlir] move if-condition propagation to a standalone pass (PR #150278)

[Mehdi Amini]

joker-eph wrote:

> That looks quite unfortunate to me to lose as a canonicalization (I don't understand your comment about locality: this does not seem like an issue to me at all actually).
> 
> Can you expand on the source of the slow-down (or speed-up)? Can you also provide a synthetic test cases to illustrate the problem?

@ftynse ping on post-merge comments!
Seems to me that this was hastily merged, which is OK only because we expect post-merge comments to be addressed promptly (or the PR reverted while we discuss).


https://github.com/llvm/llvm-project/pull/150278