[llvm-branch-commits] [llvm] release/21.x: [LV] Vectorize maxnum/minnum w/o fast-math flags. (#148239) #149736 (PR #150193)

[via llvm-branch-commits]

https://github.com/ayalz approved this pull request.

LGTM, thanks!

Independent of cherry picking: may be good to provide an example demonstrating why we revert to scalar loop when encountering NaN's, rather than try to fix the final horizontal reduction as done for cmp/sel discrepancy of +-0 by recording first/last argmax/argmin (or sign of first/last zero).

https://github.com/llvm/llvm-project/pull/150193