[llvm] [SLP] Inefficient cost-modelling and codegen for reductions with slp-… (PR #197875)

[Sushant Gokhale via llvm-commits]

================
@@ -30379,19 +30402,13 @@ class HorizontalReduction {
           if (auto *VecTy = dyn_cast<FixedVectorType>(ScalarTy)) {
             assert(SLPReVec && "FixedVectorType is not expected.");
             unsigned ScalarTyNumElements = VecTy->getNumElements();
-            for (unsigned I : seq<unsigned>(ReducedVals.size())) {
-              VectorCost += TTI->getShuffleCost(
-                  TTI::SK_PermuteSingleSrc,
-                  FixedVectorType::get(VecTy->getScalarType(),
-                                       ReducedVals.size()),
-                  VectorTy,
-                  createStrideMask(I, ScalarTyNumElements, ReducedVals.size()));
-              VectorCost += TTI->getArithmeticReductionCost(RdxOpcode, VecTy,
-                                                            FMF, CostKind);
+            for (unsigned I : seq<unsigned>(ReducedVals.size() - 1)) {
+              VectorCost +=
+                  ::getShuffleCost(*TTI, TTI::SK_ExtractSubvector, VectorTy, {},
+                                   CostKind, I * ScalarTyNumElements, VecTy);
+              VectorCost +=
+                  TTI->getArithmeticInstrCost(RdxOpcode, VecTy, CostKind);
----------------
sushgokh wrote:

done

https://github.com/llvm/llvm-project/pull/197875