[llvm] [SLP] More accurately cost RISCV scalar splats (PR #213104)

[Ryan Buchner via llvm-commits]

================
@@ -673,12 +673,14 @@ InstructionCost RISCVTTIImpl::getSlideCost(FixedVectorType *Tp,
   return FirstSlideCost + SecondSlideCost + MaskCost;
 }
 
-InstructionCost
-RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind, VectorType *DstTy,
-                             VectorType *SrcTy, ArrayRef<int> Mask,
-                             TTI::TargetCostKind CostKind, int Index,
-                             VectorType *SubTp, ArrayRef<const Value *> Args,
-                             const Instruction *CxtI) const {
+InstructionCost RISCVTTIImpl::getShuffleCost(
+    TTI::ShuffleKind Kind, VectorType *DstTy, VectorType *SrcTy,
+    ArrayRef<int> Mask, TTI::TargetCostKind CostKind, int Index,
+    VectorType *SubTp, ArrayRef<const Value *> Args, const Instruction *CxtI,
+    TTI::VectorInstrContext VIC) const {
+  if (VIC == TTI::VectorInstrContext::SplatOpFolded)
+    return TTI::TCC_Free;
----------------
bababuck wrote:

I had looked into that but deferred from it since `SplatOpFolded` should only be set in scenarios when `Kind == SK_Broadcast`. Could add as an assertion to confirm that.

https://github.com/llvm/llvm-project/pull/213104