[SLP]Include cost of the reshuffling for same nodes with resizing.

Need to account reshuffling, required for the reused elements in the
buildvector nodes, which are copies (perfect match) of other nodes, but
include reused elements.

Differential Revision: https://reviews.llvm.org/D149966

Author: alexey-bataev

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -7006,6 +7006,17 @@ class BoUpSLP::ShuffleCostEstimator : public BaseShuffleAnalysis {
         const TreeEntry *E2 = P2.get<const TreeEntry *>();
         if (E->Scalars.size() == E2->Scalars.size())
           CommonVF = VF = E->Scalars.size();
+      } else {
+        // P2 is empty, check that we have same node + reshuffle (if any).
+        if (E->Scalars.size() == Mask.size() && VF != Mask.size()) {
+          VF = E->Scalars.size();
+          SmallVector<int> CommonMask(Mask.begin(), Mask.end());
+          ::addMask(CommonMask, E->getCommonMask());
+          V1 = Constant::getNullValue(
+              FixedVectorType::get(E->Scalars.front()->getType(), VF));
+          return BaseShuffleAnalysis::createShuffle<InstructionCost>(
+              V1, nullptr, CommonMask, Builder);
+        }
       }
       V1 = Constant::getNullValue(
           FixedVectorType::get(E->Scalars.front()->getType(), VF));
@@ -7286,7 +7297,17 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
             dbgs()
             << "SLP: perfect diamond match for gather bundle that starts with "
             << *VL.front() << ".\n");
-        return 0;
+        // Restore the mask for previous partially matched values.
+        for (auto [I, V] : enumerate(E->Scalars)) {
+          if (isa<PoisonValue>(V)) {
+            Mask[I] = PoisonMaskElem;
+            continue;
+          }
+          if (Mask[I] == PoisonMaskElem)
+            Mask[I] = Entries.front()->findLaneForValue(V);
+        }
+        Estimator.add(Entries.front(), Mask);
+        return Estimator.finalize(E->ReuseShuffleIndices);
       }
       if (!Resized) {
         unsigned VF1 = Entries.front()->getVectorFactor();

llvm/test/Transforms/SLPVectorizer/X86/crash_clear_undefs.ll

@@ -9,7 +9,7 @@ target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16
 ; YAML-NEXT:  Function:        foo
 ; YAML-NEXT:  Args:
 ; YAML-NEXT:    - String:          'SLP vectorized with cost '
-; YAML-NEXT:    - Cost:            '-6'
+; YAML-NEXT:    - Cost:            '-5'
 ; YAML-NEXT:    - String:          ' and with tree size '
 ; YAML-NEXT:    - TreeSize:        '10'
 ; YAML-NEXT:  ...