[llvm] [LV] Vectorize fmin/fmax reductions (PR #198300)

[Nashe Mncube via llvm-commits]

================
@@ -2177,6 +2177,18 @@ bool VPlanTransforms::handleMultiUseReductions(VPlan &Plan,
         return Pred == CmpInst::ICMP_SLE || Pred == CmpInst::ICMP_SLT;
       case RecurKind::SMin:
         return Pred == CmpInst::ICMP_SGE || Pred == CmpInst::ICMP_SGT;
+      case RecurKind::FMax:
+      case RecurKind::FMaximum:
+      case RecurKind::FMaximumNum:
+        return Pred == CmpInst::FCMP_OLE || Pred == CmpInst::FCMP_OLT;
+      case RecurKind::FMin:
+      case RecurKind::FMinimum:
----------------
nasherm wrote:

Thanks for pointing this out. My most recent patch should resolve this

https://github.com/llvm/llvm-project/pull/198300