Clean up usages of asserting vector getters in Type

[ctetreau]

Summary

Remove usages of asserting vector getters in Type in preparation for the
VectorType refactor. The existence of these functions complicates the
refactor while adding little value.

[ctetreau]

Update to mention that this change is NFC

[ctetreau]

update commit message, rebase

[sdesmalen]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

973

This code calls cast<VectorType>(Tp) several times, so it's probably worth putting that in a separate VectorType *VTp.
A quick scan through the codebase leads me to believe that Tp is always a VectorType.

2869

cast<VectorType>(ValTy) could also fit into a separate variable. I expect ValTy is always of type VectorType.

[ctetreau]

rebase, address code review issues

[ctetreau]

add FIXME

[efriedma]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

The type has to be a vector: it's the type of a vector reduction operation. You can just cast<> here.

It would probably be worth going through at some point and making the various TTI cost hooks that require a vector take a VectorType instead of a Type. But that's out of scope here, of course.

[ctetreau]

address code review issues

[ctetreau]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

That is the endgame that I'm hoping for, but probably not something I can personally justify doing on the company's dime.

[RKSimon]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

In which case, please can you raise a bug listing the functions that need changing to VectorType?

[ctetreau]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

This issue is pervasive throughout the codebase. From what I have seen, most functions that deal in Type objects take and return pointers to the base Type class. There are many functions that immediately cast to VectorType or assert isVectorTy(), but there are many more that logically should only deal in VectorType, but instead operate on the base Type (either by casting to VectorType only in certain branches or by passing these pointers to other functions that logically only work on VectorType) Even within this module, there are many such functions. Some are obvious, many are not.

It would be a tremendous amount of work to even analyze which functions should operate on on Type subclasses rather than base objects. I do not have the bandwidth to do this analysis, and a bug that basically says "There are functions in [module] that should take VectorType instead of Type as an argument" is not a helpful bug.

These functions that I am removing are serving as a crutch for this problem. My hope is that forcing the cast will cause more code to just cast once, or take derived objects as function arguments.

[ctetreau]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

efriedma and I discussed it offline, and he agrees that at least opening the "TargetTransformInfoImpl.h has a bunch of functions that take and return base type that should take and return a vector type" bug is a useful thing to do. This file is a particularly egregious offender on this regard.

I'm in the process of getting an account on bugzilla. I will open the bug when that is completed. In the mean time, I would appreciate if we could get this patch merged as it is blocking my work.

[ctetreau]

rebase

[ctetreau]

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

2869

I've opened https://bugs.llvm.org/show_bug.cgi?id=45562 for this issue

[ctetreau]

Would it be possible to get this code reviewed soon please? This is the last commit that needs to be merged before the commit that removes the getters can be merged.

[efriedma]

LGTM

[ctetreau]

rebase

llvm/lib/Target/X86/X86InterleavedAccess.cpp

public:
   /// instructions/intrinsics.
   bool lowerIntoOptimizedSequence();
 };
 
 } // end anonymous namespace
 
 bool X86InterleavedAccessGroup::isSupported() const {
   VectorType *ShuffleVecTy = Shuffles[0]->getType();
-  Type *ShuffleEltTy = ShuffleVecTy->getVectorElementType();
+  Type *ShuffleEltTy = ShuffleVecTy->getElementType();
   unsigned ShuffleElemSize = DL.getTypeSizeInBits(ShuffleEltTy);
   unsigned WideInstSize;
 
   // Currently, lowering is supported for the following vectors:
   // Stride 4:
   //    1. Store and load of 4-element vectors of 64 bits on AVX.
   //    2. Store of 16/32-element vectors of 8 bits on AVX.
   // Stride 3:
if (auto *SVI = dyn_cast<ShuffleVectorInst>(VecInst)) {
     Value *Op0 = SVI->getOperand(0);
     Value *Op1 = SVI->getOperand(1);
 
     // Generate N(= NumSubVectors) shuffles of T(= SubVecTy) type.
     for (unsigned i = 0; i < NumSubVectors; ++i)
       DecomposedVectors.push_back(
           cast<ShuffleVectorInst>(Builder.CreateShuffleVector(
               Op0, Op1,
-              createSequentialMask(Indices[i], SubVecTy->getVectorNumElements(),
+              createSequentialMask(Indices[i], SubVecTy->getNumElements(),
                                    0))));
     return;
   }
 
   // Decompose the load instruction.
   LoadInst *LI = cast<LoadInst>(VecInst);
   Type *VecBaseTy, *VecBasePtrTy;
   Value *VecBasePtr;
bool X86InterleavedAccessGroup::lowerIntoOptimizedSequence() {
   SmallVector<Instruction *, 4> DecomposedVectors;
   SmallVector<Value *, 4> TransposedVectors;
   VectorType *ShuffleTy = Shuffles[0]->getType();
 
   if (isa<LoadInst>(Inst)) {
     // Try to generate target-sized register(/instruction).
     decompose(Inst, Factor, ShuffleTy, DecomposedVectors);
 
-    Type *ShuffleEltTy = Inst->getType();
-    unsigned NumSubVecElems = ShuffleEltTy->getVectorNumElements() / Factor;
+    auto *ShuffleEltTy = cast<VectorType>(Inst->getType());
+    unsigned NumSubVecElems = ShuffleEltTy->getNumElements() / Factor;
     // Perform matrix-transposition in order to compute interleaved
     // results by generating some sort of (optimized) target-specific
     // instructions.
 
     switch (NumSubVecElems) {
     default:
       return false;
     case 4:
if (isa<LoadInst>(Inst)) {
     // Now replace the unoptimized-interleaved-vectors with the
     // transposed-interleaved vectors.
     for (unsigned i = 0, e = Shuffles.size(); i < e; ++i)
       Shuffles[i]->replaceAllUsesWith(TransposedVectors[Indices[i]]);
 
     return true;
   }
 
-  Type *ShuffleEltTy = ShuffleTy->getVectorElementType();
-  unsigned NumSubVecElems = ShuffleTy->getVectorNumElements() / Factor;
+  Type *ShuffleEltTy = ShuffleTy->getElementType();
+  unsigned NumSubVecElems = ShuffleTy->getNumElements() / Factor;
 
   // Lower the interleaved stores:
   //   1. Decompose the interleaved wide shuffle into individual shuffle
   //   vectors.
   decompose(Shuffles[0], Factor, VectorType::get(ShuffleEltTy, NumSubVecElems),
             DecomposedVectors);
 
   //   2. Transpose the interleaved-vectors into vectors of contiguous

 }
 
 bool X86TargetLowering::lowerInterleavedStore(StoreInst *SI,
                                               ShuffleVectorInst *SVI,
                                               unsigned Factor) const {
   assert(Factor >= 2 && Factor <= getMaxSupportedInterleaveFactor() &&
          "Invalid interleave factor");
 
-  assert(SVI->getType()->getVectorNumElements() % Factor == 0 &&
+  assert(SVI->getType()->getNumElements() % Factor == 0 &&
          "Invalid interleaved store");
 
   // Holds the indices of SVI that correspond to the starting index of each
   // interleaved shuffle.
   SmallVector<unsigned, 4> Indices;
   auto Mask = SVI->getShuffleMask();
   for (unsigned i = 0; i < Factor; i++)
     Indices.push_back(Mask[i]);

llvm/lib/Target/X86/X86PartialReduction.cpp

case Instruction::FMul:
     if (auto *FPOp = dyn_cast<FPMathOperator>(&BO))
       if (FPOp->getFastMathFlags().isFast())
         break;
     LLVM_FALLTHROUGH;
   default:
     return false;
   }
 
-  unsigned ElemNum = BO.getType()->getVectorNumElements();
+  unsigned ElemNum = cast<VectorType>(BO.getType())->getNumElements();
   // Ensure the reduction size is a power of 2.
   if (!isPowerOf2_32(ElemNum))
     return false;
 
   unsigned ElemNumToReduce = ElemNum;
 
   // Do DFS search on the def-use chain from the given instruction. We only
   // allow four kinds of operations during the search until we reach the
for (const auto *U : User->users()) {
             return false;
         UsersToVisit.push_back(U);
       } else if (auto *ShufInst = dyn_cast<ShuffleVectorInst>(U)) {
         // Detect the following pattern: A ShuffleVector instruction together
         // with a reduction that do partial reduction on the first and second
         // ElemNumToReduce / 2 elements, and store the result in
         // ElemNumToReduce / 2 elements in another vector.
 
-        unsigned ResultElements = ShufInst->getType()->getVectorNumElements();
+        unsigned ResultElements = ShufInst->getType()->getNumElements();
         if (ResultElements < ElemNum)
           return false;
 
         if (ElemNumToReduce == 1)
           return false;
         if (!isa<UndefValue>(U->getOperand(1)))
           return false;
         for (unsigned i = 0; i < ElemNumToReduce / 2; ++i)
bool X86PartialReduction::tryMAddReplacement(Value *Op, BinaryOperator *Add) {
   };
 
   // Both Ops need to be shrinkable.
   if (!canShrinkOp(LHS) && !canShrinkOp(RHS))
     return false;
 
   IRBuilder<> Builder(Add);
 
-  Type *MulTy = Op->getType();
-  unsigned NumElts = MulTy->getVectorNumElements();
+  auto *MulTy = cast<VectorType>(Op->getType());
+  unsigned NumElts = MulTy->getNumElements();
 
   // Extract even elements and odd elements and add them together. This will
   // be pattern matched by SelectionDAG to pmaddwd. This instruction will be
   // half the original width.
   SmallVector<int, 16> EvenMask(NumElts / 2);
   SmallVector<int, 16> OddMask(NumElts / 2);
   for (int i = 0, e = NumElts / 2; i != e; ++i) {
     EvenMask[i] = i * 2;

 }
 
 // Try to replace operans of this add with pmaddwd patterns.
 bool X86PartialReduction::tryMAddPattern(BinaryOperator *BO) {
   if (!ST->hasSSE2())
     return false;
 
   // Need at least 8 elements.
-  if (BO->getType()->getVectorNumElements() < 8)
+  if (cast<VectorType>(BO->getType())->getNumElements() < 8)
     return false;
 
   // Element type should be i32.
-  if (!BO->getType()->getVectorElementType()->isIntegerTy(32))
+  if (!cast<VectorType>(BO->getType())->getElementType()->isIntegerTy(32))
     return false;
 
   bool Changed = false;
   Changed |= tryMAddReplacement(BO->getOperand(0), BO);
   Changed |= tryMAddReplacement(BO->getOperand(1), BO);
   return Changed;
 }
 
bool X86PartialReduction::trySADReplacement(Value *Op, BinaryOperator *Add) {
   // Need a subtract of two values.
   auto *Sub = dyn_cast<BinaryOperator>(LHS);
   if (!Sub || Sub->getOpcode() != Instruction::Sub)
     return false;
 
   // Look for zero extend from i8.
   auto getZeroExtendedVal = [](Value *Op) -> Value * {
     if (auto *ZExt = dyn_cast<ZExtInst>(Op))
-      if (ZExt->getOperand(0)->getType()->getVectorElementType()->isIntegerTy(8))
+      if (cast<VectorType>(ZExt->getOperand(0)->getType())
+              ->getElementType()
+              ->isIntegerTy(8))
         return ZExt->getOperand(0);
 
     return nullptr;
   };
 
   // Both operands of the subtract should be extends from vXi8.
   Value *Op0 = getZeroExtendedVal(Sub->getOperand(0));
   Value *Op1 = getZeroExtendedVal(Sub->getOperand(1));
   if (!Op0 || !Op1)
     return false;
 
   IRBuilder<> Builder(Add);
 
-  Type *OpTy = Op->getType();
-  unsigned NumElts = OpTy->getVectorNumElements();
+  auto *OpTy = cast<VectorType>(Op->getType());
+  unsigned NumElts = OpTy->getNumElements();
 
   unsigned IntrinsicNumElts;
   Intrinsic::ID IID;
   if (ST->hasBWI() && NumElts >= 64) {
     IID = Intrinsic::x86_avx512_psad_bw_512;
     IntrinsicNumElts = 64;
   } else if (ST->hasAVX2() && NumElts >= 32) {
     IID = Intrinsic::x86_avx2_psad_bw;
for (unsigned i = 0; i != NumSplits; ++i) {
     Value *ExtractOp1 = Builder.CreateShuffleVector(Op1, Op0, ExtractMask);
     Ops[i] = Builder.CreateCall(PSADBWFn, {ExtractOp0, ExtractOp1});
     Ops[i] = Builder.CreateBitCast(Ops[i], I32Ty);
   }
 
   assert(isPowerOf2_32(NumSplits) && "Expected power of 2 splits");
   unsigned Stages = Log2_32(NumSplits);
   for (unsigned s = Stages; s > 0; --s) {
-    unsigned NumConcatElts = Ops[0]->getType()->getVectorNumElements() * 2;
+    unsigned NumConcatElts =
+        cast<VectorType>(Ops[0]->getType())->getNumElements() * 2;
     for (unsigned i = 0; i != 1U << (s - 1); ++i) {
       SmallVector<int, 64> ConcatMask(NumConcatElts);
       std::iota(ConcatMask.begin(), ConcatMask.end(), 0);
       Ops[i] = Builder.CreateShuffleVector(Ops[i*2], Ops[i*2+1], ConcatMask);
     }
   }
 
   // At this point the final value should be in Ops[0]. Now we need to adjust
   // it to the final original type.
-  NumElts = OpTy->getVectorNumElements();
+  NumElts = cast<VectorType>(OpTy)->getNumElements();
   if (NumElts == 2) {
     // Extract down to 2 elements.
     Ops[0] = Builder.CreateShuffleVector(Ops[0], Ops[0], ArrayRef<int>{0, 1});
   } else if (NumElts >= 8) {
     SmallVector<int, 32> ConcatMask(NumElts);
-    unsigned SubElts = Ops[0]->getType()->getVectorNumElements();
+    unsigned SubElts = cast<VectorType>(Ops[0]->getType())->getNumElements();
     for (unsigned i = 0; i != SubElts; ++i)
       ConcatMask[i] = i;
     for (unsigned i = SubElts; i != NumElts; ++i)
       ConcatMask[i] = (i % SubElts) + SubElts;
 
     Value *Zero = Constant::getNullValue(Ops[0]->getType());
     Ops[0] = Builder.CreateShuffleVector(Ops[0], Zero, ConcatMask);
   }
 
   // Replaces the uses of Op in Add with the new sequence.
   Add->replaceUsesOfWith(Op, Ops[0]);
   Add->setHasNoSignedWrap(false);
   Add->setHasNoUnsignedWrap(false);
 
   return false;
 }
 
 bool X86PartialReduction::trySADPattern(BinaryOperator *BO) {
   if (!ST->hasSSE2())
     return false;
 
   // TODO: There's nothing special about i32, any integer type above i16 should
   // work just as well.
-  if (!BO->getType()->getVectorElementType()->isIntegerTy(32))
+  if (!cast<VectorType>(BO->getType())->getElementType()->isIntegerTy(32))
     return false;
 
   bool Changed = false;
   Changed |= trySADReplacement(BO->getOperand(0), BO);
   Changed |= trySADReplacement(BO->getOperand(1), BO);
   return Changed;
 }
 

llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp

static bool extractConstantMask(const Constant *C, unsigned MaskEltSizeInBits,
   // bit representation.
   // e.g. the following take up the same space in the constant pool:
   //   i128 -170141183420855150465331762880109871104
   //
   //   <2 x i64> <i64 -9223372034707292160, i64 -9223372034707292160>
   //
   //   <4 x i32> <i32 -2147483648, i32 -2147483648,
   //              i32 -2147483648, i32 -2147483648>
-  Type *CstTy = C->getType();
-  if (!CstTy->isVectorTy())
+  auto *CstTy = dyn_cast<VectorType>(C->getType());
+  if (!CstTy)
     return false;
 
-  Type *CstEltTy = CstTy->getVectorElementType();
+  Type *CstEltTy = CstTy->getElementType();
   if (!CstEltTy->isIntegerTy())
     return false;
 
   unsigned CstSizeInBits = CstTy->getPrimitiveSizeInBits();
   unsigned CstEltSizeInBits = CstTy->getScalarSizeInBits();
-  unsigned NumCstElts = CstTy->getVectorNumElements();
+  unsigned NumCstElts = CstTy->getNumElements();
 
   assert((CstSizeInBits % MaskEltSizeInBits) == 0 &&
          "Unaligned shuffle mask size");
 
   unsigned NumMaskElts = CstSizeInBits / MaskEltSizeInBits;
   UndefElts = APInt(NumMaskElts, 0);
   RawMask.resize(NumMaskElts, 0);
 

llvm/lib/Target/X86/X86TargetTransformInfo.cpp

int ScalarCost = getArithmeticInstrCost(
         TargetTransformInfo::OP_None, TargetTransformInfo::OP_None);
     return 20 * LT.first * LT.second.getVectorNumElements() * ScalarCost;
   }
 
   // Fallback to the default implementation.
   return BaseT::getArithmeticInstrCost(Opcode, Ty, Op1Info, Op2Info);
 }
 
-int X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *Tp, int Index,
+int X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *BaseTp, int Index,
                                Type *SubTp) {
+  auto *Tp = cast<VectorType>(BaseTp);
   // 64-bit packed float vectors (v2f32) are widened to type v4f32.
   // 64-bit packed integer vectors (v2i32) are widened to type v4i32.
   std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, Tp);
 
   // Treat Transpose as 2-op shuffles - there's no difference in lowering.
   if (Kind == TTI::SK_Transpose)
     Kind = TTI::SK_PermuteTwoSrc;
 
if (SubLT.second.isVector()) {
       if ((Index % NumSubElts) == 0 && (NumElts % NumSubElts) == 0)
         return SubLT.first;
       // Handle some cases for widening legalization. For now we only handle
       // cases where the original subvector was naturally aligned and evenly
       // fit in its legalized subvector type.
       // FIXME: Remove some of the alignment restrictions.
       // FIXME: We can use permq for 64-bit or larger extracts from 256-bit
       // vectors.
-      int OrigSubElts = SubTp->getVectorNumElements();
-      if (NumSubElts > OrigSubElts &&
-          (Index % OrigSubElts) == 0 && (NumSubElts % OrigSubElts) == 0 &&
+      int OrigSubElts = cast<VectorType>(SubTp)->getNumElements();
+      if (NumSubElts > OrigSubElts && (Index % OrigSubElts) == 0 &&
+          (NumSubElts % OrigSubElts) == 0 &&
           LT.second.getVectorElementType() ==
               SubLT.second.getVectorElementType() &&
           LT.second.getVectorElementType().getSizeInBits() ==
-            Tp->getVectorElementType()->getPrimitiveSizeInBits()) {
+              Tp->getElementType()->getPrimitiveSizeInBits()) {
         assert(NumElts >= NumSubElts && NumElts > OrigSubElts &&
                "Unexpected number of elements!");
-        Type *VecTy = VectorType::get(Tp->getVectorElementType(),
+        Type *VecTy = VectorType::get(Tp->getElementType(),
                                       LT.second.getVectorNumElements());
-        Type *SubTy = VectorType::get(Tp->getVectorElementType(),
+        Type *SubTy = VectorType::get(Tp->getElementType(),
                                       SubLT.second.getVectorNumElements());
         int ExtractIndex = alignDown((Index % NumElts), NumSubElts);
         int ExtractCost = getShuffleCost(TTI::SK_ExtractSubvector, VecTy,
                                          ExtractIndex, SubTy);
 
         // If the original size is 32-bits or more, we can use pshufd. Otherwise
         // if we have SSSE3 we can use pshufb.
         if (SubTp->getPrimitiveSizeInBits() >= 32 || ST->hasSSSE3())
int X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *BaseTp, int Index,
 
   // We are going to permute multiple sources and the result will be in multiple
   // destinations. Providing an accurate cost only for splits where the element
   // type remains the same.
   if (Kind == TTI::SK_PermuteSingleSrc && LT.first != 1) {
     MVT LegalVT = LT.second;
     if (LegalVT.isVector() &&
         LegalVT.getVectorElementType().getSizeInBits() ==
-            Tp->getVectorElementType()->getPrimitiveSizeInBits() &&
-        LegalVT.getVectorNumElements() < Tp->getVectorNumElements()) {
+            Tp->getElementType()->getPrimitiveSizeInBits() &&
+        LegalVT.getVectorNumElements() < Tp->getNumElements()) {
 
       unsigned VecTySize = DL.getTypeStoreSize(Tp);
       unsigned LegalVTSize = LegalVT.getStoreSize();
       // Number of source vectors after legalization:
       unsigned NumOfSrcs = (VecTySize + LegalVTSize - 1) / LegalVTSize;
       // Number of destination vectors after legalization:
       unsigned NumOfDests = LT.first;
 
-      Type *SingleOpTy = VectorType::get(Tp->getVectorElementType(),
-                                         LegalVT.getVectorNumElements());
+      Type *SingleOpTy =
+          VectorType::get(Tp->getElementType(), LegalVT.getVectorNumElements());
 
       unsigned NumOfShuffles = (NumOfSrcs - 1) * NumOfDests;
       return NumOfShuffles *
              getShuffleCost(TTI::SK_PermuteTwoSrc, SingleOpTy, 0, nullptr);
     }
 
     return BaseT::getShuffleCost(Kind, Tp, Index, SubTp);
   }
unsigned X86TTIImpl::getScalarizationOverhead(Type *Ty, bool Insert,
   return BaseT::getScalarizationOverhead(Ty, Insert, Extract);
 }
 
 int X86TTIImpl::getMemoryOpCost(unsigned Opcode, Type *Src,
                                 MaybeAlign Alignment, unsigned AddressSpace,
                                 const Instruction *I) {
   // Handle non-power-of-two vectors such as <3 x float>
   if (VectorType *VTy = dyn_cast<VectorType>(Src)) {
-    unsigned NumElem = VTy->getVectorNumElements();
+    unsigned NumElem = VTy->getNumElements();
 
     // Handle a few common cases:
     // <3 x float>
     if (NumElem == 3 && VTy->getScalarSizeInBits() == 32)
       // Cost = 64 bit store + extract + 32 bit store.
       return 3;
 
     // <3 x double>
int X86TTIImpl::getMaskedMemoryOpCost(unsigned Opcode, Type *SrcTy,
   bool IsLoad = (Instruction::Load == Opcode);
   bool IsStore = (Instruction::Store == Opcode);
 
   VectorType *SrcVTy = dyn_cast<VectorType>(SrcTy);
   if (!SrcVTy)
     // To calculate scalar take the regular cost, without mask
     return getMemoryOpCost(Opcode, SrcTy, MaybeAlign(Alignment), AddressSpace);
 
-  unsigned NumElem = SrcVTy->getVectorNumElements();
+  unsigned NumElem = SrcVTy->getNumElements();
   VectorType *MaskTy =
       VectorType::get(Type::getInt8Ty(SrcVTy->getContext()), NumElem);
   if ((IsLoad && !isLegalMaskedLoad(SrcVTy, MaybeAlign(Alignment))) ||
       (IsStore && !isLegalMaskedStore(SrcVTy, MaybeAlign(Alignment))) ||
       !isPowerOf2_32(NumElem)) {
     // Scalarization
     int MaskSplitCost = getScalarizationOverhead(MaskTy, false, true);
     int ScalarCompareCost = getCmpSelInstrCost(
int X86TTIImpl::getMaskedMemoryOpCost(unsigned Opcode, Type *SrcTy,
   int Cost = 0;
   if (VT.isSimple() && LT.second != VT.getSimpleVT() &&
       LT.second.getVectorNumElements() == NumElem)
     // Promotion requires expand/truncate for data and a shuffle for mask.
     Cost += getShuffleCost(TTI::SK_PermuteTwoSrc, SrcVTy, 0, nullptr) +
             getShuffleCost(TTI::SK_PermuteTwoSrc, MaskTy, 0, nullptr);
 
   else if (LT.second.getVectorNumElements() > NumElem) {
-    VectorType *NewMaskTy = VectorType::get(MaskTy->getVectorElementType(),
+    VectorType *NewMaskTy = VectorType::get(MaskTy->getElementType(),
                                             LT.second.getVectorNumElements());
     // Expanding requires fill mask with zeroes
     Cost += getShuffleCost(TTI::SK_InsertSubvector, NewMaskTy, 0, MaskTy);
   }
 
   // Pre-AVX512 - each maskmov load costs 2 + store costs ~8.
   if (!ST->hasAVX512())
     return Cost + LT.first * (IsLoad ? 2 : 8);
if (ST->hasSSE2())
       if (const auto *Entry = CostTableLookup(SSE2CostTblNoPairWise, ISD, MTy))
         return Entry->Cost;
   }
 
   std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, ValTy);
 
   MVT MTy = LT.second;
 
+  auto *ValVTy = cast<VectorType>(ValTy);
+
   unsigned ArithmeticCost = 0;
   if (LT.first != 1 && MTy.isVector() &&
-      MTy.getVectorNumElements() < ValTy->getVectorNumElements()) {
+      MTy.getVectorNumElements() < ValVTy->getNumElements()) {
     // Type needs to be split. We need LT.first - 1 arithmetic ops.
-    Type *SingleOpTy = VectorType::get(ValTy->getVectorElementType(),
-                                       MTy.getVectorNumElements());
+    Type *SingleOpTy =
+        VectorType::get(ValVTy->getElementType(), MTy.getVectorNumElements());
     ArithmeticCost = getArithmeticInstrCost(Opcode, SingleOpTy);
     ArithmeticCost *= LT.first - 1;
   }
 
   if (ST->isSLM())
     if (const auto *Entry = CostTableLookup(SLMCostTblNoPairWise, ISD, MTy))
       return ArithmeticCost + Entry->Cost;
 
static const CostTblEntry SSE2BoolReduction[] = {
     { ISD::AND,  MVT::v16i8,   2 }, // pmovmskb + cmp
     { ISD::OR,   MVT::v2i64,   2 }, // movmskpd + cmp
     { ISD::OR,   MVT::v4i32,   2 }, // movmskps + cmp
     { ISD::OR,   MVT::v8i16,   2 }, // pmovmskb + cmp
     { ISD::OR,   MVT::v16i8,   2 }, // pmovmskb + cmp
   };
 
   // Handle bool allof/anyof patterns.
-  if (ValTy->getVectorElementType()->isIntegerTy(1)) {
+  if (ValVTy->getElementType()->isIntegerTy(1)) {
     unsigned ArithmeticCost = 0;
     if (LT.first != 1 && MTy.isVector() &&
-        MTy.getVectorNumElements() < ValTy->getVectorNumElements()) {
+        MTy.getVectorNumElements() < ValVTy->getNumElements()) {
       // Type needs to be split. We need LT.first - 1 arithmetic ops.
-      Type *SingleOpTy = VectorType::get(ValTy->getVectorElementType(),
-                                         MTy.getVectorNumElements());
+      Type *SingleOpTy =
+          VectorType::get(ValVTy->getElementType(), MTy.getVectorNumElements());
       ArithmeticCost = getArithmeticInstrCost(Opcode, SingleOpTy);
       ArithmeticCost *= LT.first - 1;
     }
 
     if (ST->hasAVX512())
       if (const auto *Entry = CostTableLookup(AVX512BoolReduction, ISD, MTy))
         return ArithmeticCost + Entry->Cost;
     if (ST->hasAVX2())
       if (const auto *Entry = CostTableLookup(AVX2BoolReduction, ISD, MTy))
         return ArithmeticCost + Entry->Cost;
     if (ST->hasAVX())
       if (const auto *Entry = CostTableLookup(AVX1BoolReduction, ISD, MTy))
         return ArithmeticCost + Entry->Cost;
     if (ST->hasSSE2())
       if (const auto *Entry = CostTableLookup(SSE2BoolReduction, ISD, MTy))
         return ArithmeticCost + Entry->Cost;
 
-    return BaseT::getArithmeticReductionCost(Opcode, ValTy, IsPairwise);
+    return BaseT::getArithmeticReductionCost(Opcode, ValVTy, IsPairwise);
   }
 
-  unsigned NumVecElts = ValTy->getVectorNumElements();
-  unsigned ScalarSize = ValTy->getScalarSizeInBits();
+  unsigned NumVecElts = ValVTy->getNumElements();
+  unsigned ScalarSize = ValVTy->getScalarSizeInBits();
 
   // Special case power of 2 reductions where the scalar type isn't changed
   // by type legalization.
   if (!isPowerOf2_32(NumVecElts) || ScalarSize != MTy.getScalarSizeInBits())
-    return BaseT::getArithmeticReductionCost(Opcode, ValTy, IsPairwise);
+    return BaseT::getArithmeticReductionCost(Opcode, ValVTy, IsPairwise);
 
   unsigned ReductionCost = 0;
 
-  Type *Ty = ValTy;
+  auto *Ty = ValVTy;
   if (LT.first != 1 && MTy.isVector() &&
-      MTy.getVectorNumElements() < ValTy->getVectorNumElements()) {
+      MTy.getVectorNumElements() < ValVTy->getNumElements()) {
     // Type needs to be split. We need LT.first - 1 arithmetic ops.
-    Ty = VectorType::get(ValTy->getVectorElementType(),
-                         MTy.getVectorNumElements());
+    Ty = VectorType::get(ValVTy->getElementType(), MTy.getVectorNumElements());
     ReductionCost = getArithmeticInstrCost(Opcode, Ty);
     ReductionCost *= LT.first - 1;
     NumVecElts = MTy.getVectorNumElements();
   }
 
   // Now handle reduction with the legal type, taking into account size changes
   // at each level.
   while (NumVecElts > 1) {
     // Determine the size of the remaining vector we need to reduce.
     unsigned Size = NumVecElts * ScalarSize;
     NumVecElts /= 2;
     // If we're reducing from 256/512 bits, use an extract_subvector.
     if (Size > 128) {
-      Type *SubTy = VectorType::get(ValTy->getVectorElementType(), NumVecElts);
+      auto *SubTy = VectorType::get(ValVTy->getElementType(), NumVecElts);
       ReductionCost +=
           getShuffleCost(TTI::SK_ExtractSubvector, Ty, NumVecElts, SubTy);
       Ty = SubTy;
     } else if (Size == 128) {
       // Reducing from 128 bits is a permute of v2f64/v2i64.
-      Type *ShufTy;
-      if (ValTy->isFloatingPointTy())
-        ShufTy = VectorType::get(Type::getDoubleTy(ValTy->getContext()), 2);
+      VectorType *ShufTy;
+      if (ValVTy->isFloatingPointTy())
+        ShufTy = VectorType::get(Type::getDoubleTy(ValVTy->getContext()), 2);
       else
-        ShufTy = VectorType::get(Type::getInt64Ty(ValTy->getContext()), 2);
+        ShufTy = VectorType::get(Type::getInt64Ty(ValVTy->getContext()), 2);
       ReductionCost +=
           getShuffleCost(TTI::SK_PermuteSingleSrc, ShufTy, 0, nullptr);
     } else if (Size == 64) {
       // Reducing from 64 bits is a shuffle of v4f32/v4i32.
-      Type *ShufTy;
-      if (ValTy->isFloatingPointTy())
-        ShufTy = VectorType::get(Type::getFloatTy(ValTy->getContext()), 4);
+      VectorType *ShufTy;
+      if (ValVTy->isFloatingPointTy())
+        ShufTy = VectorType::get(Type::getFloatTy(ValVTy->getContext()), 4);
       else
-        ShufTy = VectorType::get(Type::getInt32Ty(ValTy->getContext()), 4);
+        ShufTy = VectorType::get(Type::getInt32Ty(ValVTy->getContext()), 4);
       ReductionCost +=
           getShuffleCost(TTI::SK_PermuteSingleSrc, ShufTy, 0, nullptr);
     } else {
       // Reducing from smaller size is a shift by immediate.
-      Type *ShiftTy = VectorType::get(
-          Type::getIntNTy(ValTy->getContext(), Size), 128 / Size);
+      auto *ShiftTy = VectorType::get(
+          Type::getIntNTy(ValVTy->getContext(), Size), 128 / Size);
       ReductionCost += getArithmeticInstrCost(
           Instruction::LShr, ShiftTy, TargetTransformInfo::OK_AnyValue,
           TargetTransformInfo::OK_UniformConstantValue,
           TargetTransformInfo::OP_None, TargetTransformInfo::OP_None);
     }
 
     // Add the arithmetic op for this level.
     ReductionCost += getArithmeticInstrCost(Opcode, Ty);
if (ST->hasSSE41())
       if (const auto *Entry = CostTableLookup(SSE41CostTblNoPairWise, ISD, MTy))
         return Entry->Cost;
 
     if (ST->hasSSE2())
       if (const auto *Entry = CostTableLookup(SSE2CostTblNoPairWise, ISD, MTy))
         return Entry->Cost;
   }
 
-  unsigned NumVecElts = ValTy->getVectorNumElements();
+  auto *ValVTy = cast<VectorType>(ValTy);
+  unsigned NumVecElts = ValVTy->getNumElements();
 
-  Type *Ty = ValTy;
+  auto *Ty = ValVTy;
   unsigned MinMaxCost = 0;
   if (LT.first != 1 && MTy.isVector() &&
-      MTy.getVectorNumElements() < ValTy->getVectorNumElements()) {
+      MTy.getVectorNumElements() < ValVTy->getNumElements()) {
     // Type needs to be split. We need LT.first - 1 operations ops.
-    Ty = VectorType::get(ValTy->getVectorElementType(),
-                         MTy.getVectorNumElements());
-    Type *SubCondTy = VectorType::get(CondTy->getVectorElementType(),
-                                      MTy.getVectorNumElements());
+    Ty = VectorType::get(ValVTy->getElementType(), MTy.getVectorNumElements());
+    Type *SubCondTy = VectorType::get(
+        cast<VectorType>(CondTy)->getElementType(), MTy.getVectorNumElements());
     MinMaxCost = getMinMaxCost(Ty, SubCondTy, IsUnsigned);
     MinMaxCost *= LT.first - 1;
     NumVecElts = MTy.getVectorNumElements();
   }
 
   if (ST->hasBWI())
     if (const auto *Entry = CostTableLookup(AVX512BWCostTblNoPairWise, ISD, MTy))
       return MinMaxCost + Entry->Cost;
int X86TTIImpl::getMinMaxReductionCost(Type *ValTy, Type *CondTy,
   if (ST->hasSSE2())
     if (const auto *Entry = CostTableLookup(SSE2CostTblNoPairWise, ISD, MTy))
       return MinMaxCost + Entry->Cost;
 
   unsigned ScalarSize = ValTy->getScalarSizeInBits();
 
   // Special case power of 2 reductions where the scalar type isn't changed
   // by type legalization.
-  if (!isPowerOf2_32(ValTy->getVectorNumElements()) ||
+  if (!isPowerOf2_32(ValVTy->getNumElements()) ||
       ScalarSize != MTy.getScalarSizeInBits())
     return BaseT::getMinMaxReductionCost(ValTy, CondTy, IsPairwise, IsUnsigned);
 
   // Now handle reduction with the legal type, taking into account size changes
   // at each level.
   while (NumVecElts > 1) {
     // Determine the size of the remaining vector we need to reduce.
     unsigned Size = NumVecElts * ScalarSize;
     NumVecElts /= 2;
     // If we're reducing from 256/512 bits, use an extract_subvector.
     if (Size > 128) {
-      Type *SubTy = VectorType::get(ValTy->getVectorElementType(), NumVecElts);
+      auto *SubTy = VectorType::get(ValVTy->getElementType(), NumVecElts);
       MinMaxCost +=
           getShuffleCost(TTI::SK_ExtractSubvector, Ty, NumVecElts, SubTy);
       Ty = SubTy;
     } else if (Size == 128) {
       // Reducing from 128 bits is a permute of v2f64/v2i64.
       Type *ShufTy;
       if (ValTy->isFloatingPointTy())
         ShufTy = VectorType::get(Type::getDoubleTy(ValTy->getContext()), 2);
if (Size > 128) {
           Type::getIntNTy(ValTy->getContext(), Size), 128 / Size);
       MinMaxCost += getArithmeticInstrCost(
           Instruction::LShr, ShiftTy, TargetTransformInfo::OK_AnyValue,
           TargetTransformInfo::OK_UniformConstantValue,
           TargetTransformInfo::OP_None, TargetTransformInfo::OP_None);
     }
 
     // Add the arithmetic op for this level.
-    Type *SubCondTy = VectorType::get(CondTy->getVectorElementType(),
-                                      Ty->getVectorNumElements());
+    auto *SubCondTy = VectorType::get(
+        cast<VectorType>(CondTy)->getElementType(), Ty->getNumElements());
     MinMaxCost += getMinMaxCost(Ty, SubCondTy, IsUnsigned);
   }
 
   // Add the final extract element to the cost.
   return MinMaxCost + getVectorInstrCost(Instruction::ExtractElement, Ty, 0);
 }
 
 /// Calculate the cost of materializing a 64-bit value. This helper
unsigned X86TTIImpl::getUserCost(const User *U,
   return BaseT::getUserCost(U, Operands);
 }
 
 // Return an average cost of Gather / Scatter instruction, maybe improved later
 int X86TTIImpl::getGSVectorCost(unsigned Opcode, Type *SrcVTy, Value *Ptr,
                                 unsigned Alignment, unsigned AddressSpace) {
 
   assert(isa<VectorType>(SrcVTy) && "Unexpected type in getGSVectorCost");
-  unsigned VF = SrcVTy->getVectorNumElements();
+  unsigned VF = cast<VectorType>(SrcVTy)->getNumElements();
 
   // Try to reduce index size from 64 bit (default for GEP)
   // to 32. It is essential for VF 16. If the index can't be reduced to 32, the
   // operation will use 16 x 64 indices which do not fit in a zmm and needs
   // to split. Also check that the base pointer is the same for all lanes,
   // and that there's at most one variable index.
   auto getIndexSizeInBits = [](Value *Ptr, const DataLayout& DL) {
     unsigned IndexSize = DL.getPointerSizeInBits();
     GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr);
     if (IndexSize < 64 || !GEP)
       return IndexSize;
 
     unsigned NumOfVarIndices = 0;
     Value *Ptrs = GEP->getPointerOperand();
     if (Ptrs->getType()->isVectorTy() && !getSplatValue(Ptrs))
       return IndexSize;
     for (unsigned i = 1; i < GEP->getNumOperands(); ++i) {
       if (isa<Constant>(GEP->getOperand(i)))
         continue;
       Type *IndxTy = GEP->getOperand(i)->getType();
-      if (IndxTy->isVectorTy())
-        IndxTy = IndxTy->getVectorElementType();
+      if (auto *IndexVTy = dyn_cast<VectorType>(IndxTy))
+        IndxTy = IndexVTy->getElementType();
       if ((IndxTy->getPrimitiveSizeInBits() == 64 &&
           !isa<SExtInst>(GEP->getOperand(i))) ||
          ++NumOfVarIndices > 1)
         return IndexSize; // 64
     }
     return (unsigned)32;
   };
 

 /// SrcVTy - The type of the data vector that should be gathered or scattered.
 /// VariableMask - The mask is non-constant at compile time.
 /// Alignment - Alignment for one element.
 /// AddressSpace - pointer[s] address space.
 ///
 int X86TTIImpl::getGSScalarCost(unsigned Opcode, Type *SrcVTy,
                                 bool VariableMask, unsigned Alignment,
                                 unsigned AddressSpace) {
-  unsigned VF = SrcVTy->getVectorNumElements();
+  unsigned VF = cast<VectorType>(SrcVTy)->getNumElements();
 
   int MaskUnpackCost = 0;
   if (VariableMask) {
     VectorType *MaskTy =
       VectorType::get(Type::getInt1Ty(SrcVTy->getContext()), VF);
     MaskUnpackCost = getScalarizationOverhead(MaskTy, false, true);
     int ScalarCompareCost =
       getCmpSelInstrCost(Instruction::ICmp, Type::getInt1Ty(SrcVTy->getContext()),

 }
 
 /// Calculate the cost of Gather / Scatter operation
 int X86TTIImpl::getGatherScatterOpCost(unsigned Opcode, Type *SrcVTy,
                                        Value *Ptr, bool VariableMask,
                                        unsigned Alignment,
                                        const Instruction *I = nullptr) {
   assert(SrcVTy->isVectorTy() && "Unexpected data type for Gather/Scatter");
-  unsigned VF = SrcVTy->getVectorNumElements();
+  unsigned VF = cast<VectorType>(SrcVTy)->getNumElements();
   PointerType *PtrTy = dyn_cast<PointerType>(Ptr->getType());
   if (!PtrTy && Ptr->getType()->isVectorTy())
-    PtrTy = dyn_cast<PointerType>(Ptr->getType()->getVectorElementType());
+    PtrTy = dyn_cast<PointerType>(
+        cast<VectorType>(Ptr->getType())->getElementType());
   assert(PtrTy && "Unexpected type for Ptr argument");
   unsigned AddressSpace = PtrTy->getAddressSpace();
 
   bool Scalarize = false;
   if ((Opcode == Instruction::Load &&
        !isLegalMaskedGather(SrcVTy, MaybeAlign(Alignment))) ||
       (Opcode == Instruction::Store &&
        !isLegalMaskedScatter(SrcVTy, MaybeAlign(Alignment))))
bool X86TTIImpl::canMacroFuseCmp() {
   return ST->hasMacroFusion() || ST->hasBranchFusion();
 }
 
 bool X86TTIImpl::isLegalMaskedLoad(Type *DataTy, MaybeAlign Alignment) {
   if (!ST->hasAVX())
     return false;
 
   // The backend can't handle a single element vector.
-  if (isa<VectorType>(DataTy) && DataTy->getVectorNumElements() == 1)
+  if (isa<VectorType>(DataTy) &&
+      cast<VectorType>(DataTy)->getNumElements() == 1)
     return false;
   Type *ScalarTy = DataTy->getScalarType();
 
   if (ScalarTy->isPointerTy())
     return true;
 
   if (ScalarTy->isFloatTy() || ScalarTy->isDoubleTy())
     return true;

 bool X86TTIImpl::isLegalMaskedExpandLoad(Type *DataTy) {
   if (!isa<VectorType>(DataTy))
     return false;
 
   if (!ST->hasAVX512())
     return false;
 
   // The backend can't handle a single element vector.
-  if (DataTy->getVectorNumElements() == 1)
+  if (cast<VectorType>(DataTy)->getNumElements() == 1)
     return false;
 
-  Type *ScalarTy = DataTy->getVectorElementType();
+  Type *ScalarTy = cast<VectorType>(DataTy)->getElementType();
 
   if (ScalarTy->isFloatTy() || ScalarTy->isDoubleTy())
     return true;
 
   if (!ScalarTy->isIntegerTy())
     return false;
 
   unsigned IntWidth = ScalarTy->getIntegerBitWidth();
bool X86TTIImpl::isLegalMaskedGather(Type *DataTy, MaybeAlign Alignment) {
   // sends a scalar type and the decision is based on the width of the
   // scalar element.
   // Later on, the cost model will estimate usage this intrinsic based on
   // the vector type.
   // The Scalarizer asks again about legality. It sends a vector type.
   // In this case we can reject non-power-of-2 vectors.
   // We also reject single element vectors as the type legalizer can't
   // scalarize it.
-  if (isa<VectorType>(DataTy)) {
-    unsigned NumElts = DataTy->getVectorNumElements();
+  if (auto *DataVTy = dyn_cast<VectorType>(DataTy)) {
+    unsigned NumElts = DataVTy->getNumElements();
     if (NumElts == 1 || !isPowerOf2_32(NumElts))
       return false;
   }
   Type *ScalarTy = DataTy->getScalarType();
   if (ScalarTy->isPointerTy())
     return true;
 
   if (ScalarTy->isFloatTy() || ScalarTy->isDoubleTy())
int X86TTIImpl::getInterleavedMemoryOpCostAVX2(unsigned Opcode, Type *VecTy,
 
   // This function can be called with VecTy=<6xi128>, Factor=3, in which case
   // the VF=2, while v2i128 is an unsupported MVT vector type
   // (see MachineValueType.h::getVectorVT()).
   if (!LegalVT.isVector())
     return BaseT::getInterleavedMemoryOpCost(Opcode, VecTy, Factor, Indices,
                                              Alignment, AddressSpace);
 
-  unsigned VF = VecTy->getVectorNumElements() / Factor;
-  Type *ScalarTy = VecTy->getVectorElementType();
+  unsigned VF = cast<VectorType>(VecTy)->getNumElements() / Factor;
+  Type *ScalarTy = cast<VectorType>(VecTy)->getElementType();
 
   // Calculate the number of memory operations (NumOfMemOps), required
   // for load/store the VecTy.
   unsigned VecTySize = DL.getTypeStoreSize(VecTy);
   unsigned LegalVTSize = LegalVT.getStoreSize();
   unsigned NumOfMemOps = (VecTySize + LegalVTSize - 1) / LegalVTSize;
 
   // Get the cost of one memory operation.
-  Type *SingleMemOpTy = VectorType::get(VecTy->getVectorElementType(),
-                                        LegalVT.getVectorNumElements());
+  Type *SingleMemOpTy =
+      VectorType::get(cast<VectorType>(VecTy)->getElementType(),
   unsigned MemOpCost = getMemoryOpCost(Opcode, SingleMemOpTy,
                                        MaybeAlign(Alignment), AddressSpace);
 
   VectorType *VT = VectorType::get(ScalarTy, VF);
   EVT ETy = TLI->getValueType(DL, VT);
   if (!ETy.isSimple())
     return BaseT::getInterleavedMemoryOpCost(Opcode, VecTy, Factor, Indices,
                                              Alignment, AddressSpace);
int X86TTIImpl::getInterleavedMemoryOpCostAVX512(unsigned Opcode, Type *VecTy,
   // Calculate the number of memory operations (NumOfMemOps), required
   // for load/store the VecTy.
   MVT LegalVT = getTLI()->getTypeLegalizationCost(DL, VecTy).second;
   unsigned VecTySize = DL.getTypeStoreSize(VecTy);
   unsigned LegalVTSize = LegalVT.getStoreSize();
   unsigned NumOfMemOps = (VecTySize + LegalVTSize - 1) / LegalVTSize;
 
   // Get the cost of one memory operation.
-  Type *SingleMemOpTy = VectorType::get(VecTy->getVectorElementType(),
-                                        LegalVT.getVectorNumElements());
+  Type *SingleMemOpTy =
+      VectorType::get(cast<VectorType>(VecTy)->getElementType(),
   unsigned MemOpCost = getMemoryOpCost(Opcode, SingleMemOpTy,
                                        MaybeAlign(Alignment), AddressSpace);
 
-  unsigned VF = VecTy->getVectorNumElements() / Factor;
+  unsigned VF = cast<VectorType>(VecTy)->getNumElements() / Factor;
   MVT VT = MVT::getVectorVT(MVT::getVT(VecTy->getScalarType()), VF);
 
   if (Opcode == Instruction::Load) {
     // The tables (AVX512InterleavedLoadTbl and AVX512InterleavedStoreTbl)
     // contain the cost of the optimized shuffle sequence that the
     // X86InterleavedAccess pass will generate.
     // The cost of loads and stores are computed separately from the table.
 
if (Opcode == Instruction::Load) {
     TTI::ShuffleKind ShuffleKind =
         (NumOfMemOps > 1) ? TTI::SK_PermuteTwoSrc : TTI::SK_PermuteSingleSrc;
 
     unsigned ShuffleCost =
         getShuffleCost(ShuffleKind, SingleMemOpTy, 0, nullptr);
 
     unsigned NumOfLoadsInInterleaveGrp =
         Indices.size() ? Indices.size() : Factor;
-    Type *ResultTy = VectorType::get(VecTy->getVectorElementType(),
-                                     VecTy->getVectorNumElements() / Factor);
+    Type *ResultTy =
+        VectorType::get(cast<VectorType>(VecTy)->getElementType(),
+                        cast<VectorType>(VecTy)->getNumElements() / Factor);
     unsigned NumOfResults =
         getTLI()->getTypeLegalizationCost(DL, ResultTy).first *
         NumOfLoadsInInterleaveGrp;
 
     // About a half of the loads may be folded in shuffles when we have only
     // one result. If we have more than one result, we do not fold loads at all.
     unsigned NumOfUnfoldedLoads =
         NumOfResults > 1 ? NumOfMemOps : NumOfMemOps / 2;

 int X86TTIImpl::getInterleavedMemoryOpCost(unsigned Opcode, Type *VecTy,
                                            unsigned Factor,
                                            ArrayRef<unsigned> Indices,
                                            unsigned Alignment,
                                            unsigned AddressSpace,
                                            bool UseMaskForCond,
                                            bool UseMaskForGaps) {
   auto isSupportedOnAVX512 = [](Type *VecTy, bool HasBW) {
-    Type *EltTy = VecTy->getVectorElementType();
+    Type *EltTy = cast<VectorType>(VecTy)->getElementType();
     if (EltTy->isFloatTy() || EltTy->isDoubleTy() || EltTy->isIntegerTy(64) ||
         EltTy->isIntegerTy(32) || EltTy->isPointerTy())
       return true;
     if (EltTy->isIntegerTy(16) || EltTy->isIntegerTy(8))
       return HasBW;
     return false;
   };
   if (ST->hasAVX512() && isSupportedOnAVX512(VecTy, ST->hasBWI()))