[llvm] [LV] Vectorize FMax via OrderedFCmpSelect w/o fast-math flags. (PR #146711)

[via llvm-commits]

================
@@ -1310,6 +1312,7 @@ Value *llvm::createSimpleReduction(IRBuilderBase &Builder, Value *Src,
   case RecurKind::SMin:
   case RecurKind::UMax:
   case RecurKind::UMin:
+  case RecurKind::OrderedFCmpSelect:
   case RecurKind::FMax:
   case RecurKind::FMin:
----------------
ayalz wrote:

Perhaps better placed after fast/simple FMax/FMin, right before complex FMaxNum/FMinNum.

https://github.com/llvm/llvm-project/pull/146711