[llvm] Reapply "[LoopUnroll] Support parallel reductions for minmax" (PR #201010)

[Madhur Amilkanthwar via llvm-commits]

================
@@ -1312,6 +1312,10 @@ Value *llvm::createMinMaxOp(IRBuilderBase &Builder, RecurKind RK, Value *Left,
   CmpInst::Predicate Pred = getMinMaxReductionPredicate(RK);
   Value *Cmp = Builder.CreateCmp(Pred, Left, Right, "rdx.minmax.cmp");
   Value *Select = Builder.CreateSelect(Cmp, Left, Right, "rdx.minmax.select");
+  // The freshly built select is not derived from any existing IR
+  // instruction; mark its branch weights as explicitly unknown.
+  if (auto *SI = dyn_cast<SelectInst>(Select))
+    setExplicitlyUnknownBranchWeightsIfProfiled(*SI, DEBUG_TYPE);
----------------
madhur13490 wrote:

@fhahn Given that #201750 is now merged, does this look good now?

https://github.com/llvm/llvm-project/pull/201010