[llvm] [VectorCombine] Scalarize bin ops and cmps with two splatted operands (PR #137786)

[Luke Lau via llvm-commits]

================
@@ -1099,29 +1110,33 @@ bool VectorCombine::scalarizeBinopOrCmp(Instruction &I) {
   // Get cost estimate for the insert element. This cost will factor into
   // both sequences.
   InstructionCost InsertCost = TTI.getVectorInstrCost(
-      Instruction::InsertElement, VecTy, CostKind, Index);
-  InstructionCost OldCost =
-      (IsConst0 ? 0 : InsertCost) + (IsConst1 ? 0 : InsertCost) + VectorOpCost;
-  InstructionCost NewCost = ScalarOpCost + InsertCost +
-                            (IsConst0 ? 0 : !Ins0->hasOneUse() * InsertCost) +
-                            (IsConst1 ? 0 : !Ins1->hasOneUse() * InsertCost);
+      Instruction::InsertElement, VecTy, CostKind, Index.value_or(0));
+  InstructionCost OldCost = (isa<Constant>(Ins0) ? 0 : InsertCost) +
----------------
lukel97 wrote:

Just noting that as far as I can tell we don't have a "splat from scalar" cost hook, which is why I originally kept it as just an insert. I think the technically correct cost would be an insert element + broadcast shuffle. But that would overcost it on RISC-V I presume?

https://github.com/llvm/llvm-project/pull/137786