[llvm-branch-commits] [llvm] [AArch64] Fold MIN/MAX(Vec[0], Vec[1]) to VECREDUCE_MIN/MAX(Vec) (PR #181162)

[Paul Walker via llvm-branch-commits]

paulwalker-arm wrote:

> I can put up both PRs if you like? And see which wins 🙂

That depends.  There should be a canonical form regardless, so if you're saying the equivalent vector.reduce.smin is already canonicalised to extracts with a scalar smin, then this PR makes sense.  If there is no such canonicalisation, then it's worth settling that debate first.

https://github.com/llvm/llvm-project/pull/181162