Reapply "[RISCV][TTI] Add shuffle costing for masked slide lowering (#128537)"

With a fix for fully undef masks.  These can't reach the lowering code, but
can reach the costing code via e.g. SLP.

This change adds the TTI costing corresponding to the recently added
isMaskedSlidePair lowering for vector shuffles. However, since the
existing costing code hadn't covered either slideup, slidedown, or the
(now removed) isElementRotate, the impact is larger in scope than just
that new lowering.

---------

Co-authored-by: Alexey Bataev <[email protected]>
Co-authored-by: Luke Lau <[email protected]>

Author: 3 people

llvm/include/llvm/Analysis/VectorUtils.h

@@ -203,6 +203,15 @@ bool getShuffleDemandedElts(int SrcWidth, ArrayRef<int> Mask,
                             const APInt &DemandedElts, APInt &DemandedLHS,
                             APInt &DemandedRHS, bool AllowUndefElts = false);
 
+/// Does this shuffle mask represent either one slide shuffle or a pair of
+/// two slide shuffles, combined with a select on some constant vector mask?
+/// A slide is a shuffle mask which shifts some set of elements up or down
+/// the vector, with all other elements being undefined.  An identity shuffle
+/// will be matched a slide by 0.  The output parameter provides the source
+/// (-1 means no source), and slide direction for each slide.
+bool isMaskedSlidePair(ArrayRef<int> Mask, int NumElts,
+                       std::array<std::pair<int, int>, 2> &SrcInfo);
+
 /// Replace each shuffle mask index with the scaled sequential indices for an
 /// equivalent mask of narrowed elements. Mask elements that are less than 0
 /// (sentinel values) are repeated in the output mask.

llvm/lib/Analysis/VectorUtils.cpp

@@ -419,6 +419,36 @@ bool llvm::getShuffleDemandedElts(int SrcWidth, ArrayRef<int> Mask,
   return true;
 }
 
+bool llvm::isMaskedSlidePair(ArrayRef<int> Mask, int NumElts,
+                             std::array<std::pair<int, int>, 2> &SrcInfo) {
+  const int SignalValue = NumElts * 2;
+  SrcInfo[0] = {-1, SignalValue};
+  SrcInfo[1] = {-1, SignalValue};
+  for (auto [i, M] : enumerate(Mask)) {
+    if (M < 0)
+      continue;
+    int Src = M >= (int)NumElts;
+    int Diff = (int)i - (M % NumElts);
+    bool Match = false;
+    for (int j = 0; j < 2; j++) {
+      auto &[SrcE, DiffE] = SrcInfo[j];
+      if (SrcE == -1) {
+        assert(DiffE == SignalValue);
+        SrcE = Src;
+        DiffE = Diff;
+      }
+      if (SrcE == Src && DiffE == Diff) {
+        Match = true;
+        break;
+      }
+    }
+    if (!Match)
+      return false;
+  }
+  // Avoid all undef masks
+  return SrcInfo[0].first != -1;
+}
+
 void llvm::narrowShuffleMaskElts(int Scale, ArrayRef<int> Mask,
                                  SmallVectorImpl<int> &ScaledMask) {
   assert(Scale > 0 && "Unexpected scaling factor");

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

@@ -4584,32 +4584,9 @@ static bool isInterleaveShuffle(ArrayRef<int> Mask, MVT VT, int &EvenSrc,
 
 /// Is this mask representing a masked combination of two slides?
 static bool isMaskedSlidePair(ArrayRef<int> Mask,
-                              std::pair<int, int> SrcInfo[2]) {
-  int NumElts = Mask.size();
-  int SignalValue = NumElts * 2;
-  SrcInfo[0] = {-1, SignalValue};
-  SrcInfo[1] = {-1, SignalValue};
-  for (unsigned i = 0; i != Mask.size(); ++i) {
-    int M = Mask[i];
-    if (M < 0)
-      continue;
-    int Src = M >= (int)NumElts;
-    int Diff = (int)i - (M % NumElts);
-    bool Match = false;
-    for (int j = 0; j < 2; j++) {
-      if (SrcInfo[j].first == -1) {
-        assert(SrcInfo[j].second == SignalValue);
-        SrcInfo[j].first = Src;
-        SrcInfo[j].second = Diff;
-      }
-      if (SrcInfo[j].first == Src && SrcInfo[j].second == Diff) {
-        Match = true;
-        break;
-      }
-    }
-    if (!Match)
-      return false;
-  }
+                              std::array<std::pair<int, int>, 2> &SrcInfo) {
+  if (!llvm::isMaskedSlidePair(Mask, Mask.size(), SrcInfo))
+    return false;
 
   // Avoid matching vselect idioms
   if (SrcInfo[0].second == 0 && SrcInfo[1].second == 0)
@@ -4625,7 +4602,8 @@ static bool isMaskedSlidePair(ArrayRef<int> Mask,
 
 // Exactly matches the semantics of a previously existing custom matcher
 // to allow migration to new matcher without changing output.
-static bool isElementRotate(std::pair<int, int> SrcInfo[2], unsigned NumElts) {
+static bool isElementRotate(std::array<std::pair<int, int>, 2> &SrcInfo,
+                            unsigned NumElts) {
   if (SrcInfo[1].first == -1)
     return true;
   return SrcInfo[0].second < 0 && SrcInfo[1].second > 0 &&
@@ -5626,10 +5604,10 @@ static SDValue lowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG,
   // without masking.  Avoid matching bit rotates (which are not also element
   // rotates) as slide pairs.  This is a performance heuristic, not a
   // functional check.
-  std::pair<int, int> SrcInfo[2];
+  std::array<std::pair<int, int>, 2> SrcInfo;
   unsigned RotateAmt;
   MVT RotateVT;
-  if (isMaskedSlidePair(Mask, SrcInfo) &&
+  if (::isMaskedSlidePair(Mask, SrcInfo) &&
       (isElementRotate(SrcInfo, NumElts) ||
        !isLegalBitRotate(Mask, VT, Subtarget, RotateVT, RotateAmt))) {
     SDValue Sources[2];
@@ -5986,10 +5964,11 @@ bool RISCVTargetLowering::isShuffleMaskLegal(ArrayRef<int> M, EVT VT) const {
   if (SVT.getScalarType() == MVT::i1)
     return false;
 
-  std::pair<int, int> SrcInfo[2];
+  std::array<std::pair<int, int>, 2> SrcInfo;
   int Dummy1, Dummy2;
   return ShuffleVectorInst::isReverseMask(M, NumElts) ||
-         (isMaskedSlidePair(M, SrcInfo) && isElementRotate(SrcInfo, NumElts)) ||
+         (::isMaskedSlidePair(M, SrcInfo) &&
+          isElementRotate(SrcInfo, NumElts)) ||
          isInterleaveShuffle(M, SVT, Dummy1, Dummy2, Subtarget);
 }
 

llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp

@@ -475,6 +475,64 @@ costShuffleViaVRegSplitting(RISCVTTIImpl &TTI, MVT LegalVT,
   return InstructionCost::getInvalid();
 }
 
+InstructionCost RISCVTTIImpl::getSlideCost(FixedVectorType *Tp,
+                                           ArrayRef<int> Mask,
+                                           TTI::TargetCostKind CostKind) {
+  // Avoid missing masks and length changing shuffles
+  if (Mask.size() <= 2 || Mask.size() != Tp->getNumElements())
+    return InstructionCost::getInvalid();
+
+  int NumElts = Tp->getNumElements();
+  std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(Tp);
+  // Avoid scalarization cases
+  if (!LT.second.isFixedLengthVector())
+    return InstructionCost::getInvalid();
+
+  // Requires moving elements between parts, which requires additional
+  // unmodeled instructions.
+  if (LT.first != 1)
+    return InstructionCost::getInvalid();
+
+  auto GetSlideOpcode = [&](int SlideAmt) {
+    assert(SlideAmt != 0);
+    bool IsVI = isUInt<5>(std::abs(SlideAmt));
+    if (SlideAmt < 0)
+      return IsVI ? RISCV::VSLIDEDOWN_VI : RISCV::VSLIDEDOWN_VX;
+    return IsVI ? RISCV::VSLIDEUP_VI : RISCV::VSLIDEUP_VX;
+  };
+
+  std::array<std::pair<int, int>, 2> SrcInfo;
+  if (!isMaskedSlidePair(Mask, NumElts, SrcInfo))
+    return InstructionCost::getInvalid();
+
+  if (SrcInfo[1].second == 0)
+    std::swap(SrcInfo[0], SrcInfo[1]);
+
+  InstructionCost FirstSlideCost = 0;
+  if (SrcInfo[0].second != 0) {
+    unsigned Opcode = GetSlideOpcode(SrcInfo[0].second);
+    FirstSlideCost = getRISCVInstructionCost(Opcode, LT.second, CostKind);
+  }
+
+  if (SrcInfo[1].first == -1)
+    return FirstSlideCost;
+
+  InstructionCost SecondSlideCost = 0;
+  if (SrcInfo[1].second != 0) {
+    unsigned Opcode = GetSlideOpcode(SrcInfo[1].second);
+    SecondSlideCost = getRISCVInstructionCost(Opcode, LT.second, CostKind);
+  } else {
+    SecondSlideCost =
+        getRISCVInstructionCost(RISCV::VMERGE_VVM, LT.second, CostKind);
+  }
+
+  auto EC = Tp->getElementCount();
+  VectorType *MaskTy =
+      VectorType::get(IntegerType::getInt1Ty(Tp->getContext()), EC);
+  InstructionCost MaskCost = getConstantPoolLoadCost(MaskTy, CostKind);
+  return FirstSlideCost + SecondSlideCost + MaskCost;
+}
+
 InstructionCost RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
                                              VectorType *Tp, ArrayRef<int> Mask,
                                              TTI::TargetCostKind CostKind,
@@ -487,8 +545,8 @@ InstructionCost RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
   // First, handle cases where having a fixed length vector enables us to
   // give a more accurate cost than falling back to generic scalable codegen.
   // TODO: Each of these cases hints at a modeling gap around scalable vectors.
-  if (ST->hasVInstructions() && isa<FixedVectorType>(Tp) &&
-      LT.second.isFixedLengthVector()) {
+  if (auto *FVTp = dyn_cast<FixedVectorType>(Tp);
+      FVTp && ST->hasVInstructions() && LT.second.isFixedLengthVector()) {
     InstructionCost VRegSplittingCost = costShuffleViaVRegSplitting(
         *this, LT.second, ST->getRealVLen(), Tp, Mask, CostKind);
     if (VRegSplittingCost.isValid())
@@ -544,6 +602,11 @@ InstructionCost RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
           return Cost;
         }
       }
+
+      if (InstructionCost SlideCost = getSlideCost(FVTp, Mask, CostKind);
+          SlideCost.isValid())
+        return SlideCost;
+
       // vrgather + cost of generating the mask constant.
       // We model this for an unknown mask with a single vrgather.
       if (LT.first == 1 && (LT.second.getScalarSizeInBits() != 8 ||
@@ -558,6 +621,11 @@ InstructionCost RISCVTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
     }
     case TTI::SK_Transpose:
     case TTI::SK_PermuteTwoSrc: {
+
+      if (InstructionCost SlideCost = getSlideCost(FVTp, Mask, CostKind);
+          SlideCost.isValid())
+        return SlideCost;
+
       // 2 x (vrgather + cost of generating the mask constant) + cost of mask
       // register for the second vrgather. We model this for an unknown
       // (shuffle) mask.

llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h

@@ -63,6 +63,12 @@ class RISCVTTIImpl : public BasicTTIImplBase<RISCVTTIImpl> {
   /// type.
   InstructionCost getConstantPoolLoadCost(Type *Ty,
                                           TTI::TargetCostKind CostKind);
+
+  /// If this shuffle can be lowered as a masked slide pair (at worst),
+  /// return a cost for it.
+  InstructionCost getSlideCost(FixedVectorType *Tp, ArrayRef<int> Mask,
+                               TTI::TargetCostKind CostKind);
+
 public:
   explicit RISCVTTIImpl(const RISCVTargetMachine *TM, const Function &F)
       : BaseT(TM, F.getDataLayout()), ST(TM->getSubtargetImpl(F)),

llvm/test/Analysis/CostModel/RISCV/shuffle-exact-vlen.ll

@@ -186,7 +186,7 @@ define void @insert_subvec() vscale_range(2,2) {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_1 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 16, i32 17, i32 18, i32 19, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_2 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_3 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_4_05 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 19, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16i32_4_05 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 19, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; CHECK-SIZE-LABEL: 'insert_subvec'
@@ -225,7 +225,7 @@ define void @insert_subvec() vscale_range(2,2) {
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_1 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 16, i32 17, i32 18, i32 19, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_2 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i32_4_3 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 16, i32 17, i32 18, i32 19>
-; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_4_05 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 19, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16i32_4_05 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 16, i32 17, i32 18, i32 19, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %v4i8_2_0 = shufflevector <4 x i8> poison, <4 x i8> poison, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
@@ -737,8 +737,8 @@ define void @multipart() vscale_range(2,2) {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16b = shufflevector <8 x i16> poison, <8 x i16> poison, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16c = shufflevector <16 x i16> poison, <16 x i16> poison, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16d = shufflevector <16 x i16> poison, <16 x i16> poison, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32a = shufflevector <4 x i32> poison, <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v32a4 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v32a = shufflevector <4 x i32> poison, <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %v32a4 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v32idrev = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 31, i32 30, i32 29, i32 28>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %v32many = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %v32many2 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 1, i32 4, i32 8, i32 12, i32 17, i32 20, i32 24, i32 28, i32 2, i32 6, i32 11, i32 14, i32 18, i32 22, i32 27, i32 30>
@@ -757,8 +757,8 @@ define void @multipart() vscale_range(2,2) {
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16b = shufflevector <8 x i16> poison, <8 x i16> poison, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16c = shufflevector <16 x i16> poison, <16 x i16> poison, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16d = shufflevector <16 x i16> poison, <16 x i16> poison, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32a = shufflevector <4 x i32> poison, <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
-; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v32a4 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v32a = shufflevector <4 x i32> poison, <4 x i32> poison, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %v32a4 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v32idrev = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 31, i32 30, i32 29, i32 28>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %v32many = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
 ; CHECK-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %v32many2 = shufflevector <16 x i32> poison, <16 x i32> poison, <16 x i32> <i32 1, i32 4, i32 8, i32 12, i32 17, i32 20, i32 24, i32 28, i32 2, i32 6, i32 11, i32 14, i32 18, i32 22, i32 27, i32 30>

llvm/test/Analysis/CostModel/RISCV/shuffle-extract_subvector.ll

@@ -19,7 +19,7 @@ define void @test_vXf64(<4 x double> %src256, <8 x double> %src512) {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; VLEN128-LABEL: 'test_vXf64'