Few bug fixes in HexagonVectorCombine

[quic_aankit]

Summary

Use getTypeAllocSize to compute difference between objects

The code was using getTypeStoreSize to calculate the difference
between consecutive objects. The calculation was incorrect due
to padding that is added between consecutive objects. The
getTypeAllocSize includes the padding amount. For example,
if the type is [19 x i8], the difference between consecutive
objects is 32 bytes, not 19 bytes.

A second case for getTypeAllocSize is needed when computing
the pointer values for the vector accesses. The calculation needs
to account for the padding as well.

Patch authored by: Brendon Cahoon <[e-mail]>

[quic_aankit]

This comment was removed by quic_aankit.

llvm/lib/Target/Hexagon/HexagonVectorCombine.cpp

public:
   Optional<APInt> getIntValue(const Value *Val) const;
   // Is V a constant 0, or a vector of 0s?
   bool isZero(const Value *Val) const;
   // Is V an undef value?
   bool isUndef(const Value *Val) const;
 
   int getSizeOf(const Value *Val) const;
   int getSizeOf(const Type *Ty) const;
+  int getAllocSizeOf(const Type *Ty) const;
   int getTypeAlignment(Type *Ty) const;
 
   VectorType *getByteVectorTy(int ScLen) const;
   Constant *getNullValue(Type *Ty) const;
   Constant *getFullValue(Type *Ty) const;
 
   Value *insertb(IRBuilder<> &Builder, Value *Dest, Value *Src, int Start,
                  int Length, int Where) const;
if (auto *II = dyn_cast<IntrinsicInst>(Val)) {
     }
   }
 
   Type *ValTy = getPayload(Val)->getType();
   if (auto *VecTy = dyn_cast<VectorType>(ValTy)) {
     int ElemCount = VecTy->getElementCount().getFixedValue();
     return HVC.getFullValue(HVC.getBoolTy(ElemCount));
   }
-  return HVC.getFullValue(HVC.getBoolTy());
+  // For scalars, return a vector <1 x i1>.
+  return HVC.getFullValue(HVC.getBoolTy(1));
 }
 
 auto AlignVectors::getPassThrough(Value *Val) const -> Value * {
   if (auto *II = dyn_cast<IntrinsicInst>(Val)) {
     if (II->getIntrinsicID() == Intrinsic::masked_load)
       return II->getArgOperand(3);
   }
   return UndefValue::get(getPayload(Val)->getType());
 }
 
 auto AlignVectors::createAdjustedPointer(IRBuilder<> &Builder, Value *Ptr,
                                          Type *ValTy, int Adjust) const
     -> Value * {
   // The adjustment is in bytes, but if it's a multiple of the type size,
   // we don't need to do pointer casts.
   auto *PtrTy = cast<PointerType>(Ptr->getType());
   if (!PtrTy->isOpaque()) {
     Type *ElemTy = PtrTy->getElementType();
-    int ElemSize = HVC.getSizeOf(ElemTy);
-    if (Adjust % ElemSize == 0) {
+    int ElemSize = HVC.getAllocSizeOf(ElemTy);
+    if (Adjust % ElemSize == 0 && Adjust != 0) {
       Value *Tmp0 =
           Builder.CreateGEP(ElemTy, Ptr, HVC.getConstInt(Adjust / ElemSize));
       return Builder.CreatePointerCast(Tmp0, ValTy->getPointerTo());
     }
   }
 
   PointerType *CharPtrTy = Type::getInt8PtrTy(HVC.F.getContext());
   Value *Tmp0 = Builder.CreatePointerCast(Ptr, CharPtrTy);
auto AlignVectors::realignGroup(const MoveGroup &Move) const -> bool {
   std::set<Instruction *> TestSet(Move.Main.begin(), Move.Main.end());
   AddrList MoveInfos;
   llvm::copy_if(
       BaseInfos, std::back_inserter(MoveInfos),
       [&TestSet](const AddrInfo &AI) { return TestSet.count(AI.Inst); });
 
   // Maximum alignment present in the whole address group.
   const AddrInfo &WithMaxAlign =
-      getMaxOf(BaseInfos, [](const AddrInfo &AI) { return AI.HaveAlign; });
+      getMaxOf(MoveInfos, [](const AddrInfo &AI) { return AI.HaveAlign; });
   Align MaxGiven = WithMaxAlign.HaveAlign;
 
   // Minimum alignment present in the move address group.
   const AddrInfo &WithMinOffset =
       getMaxOf(MoveInfos, [](const AddrInfo &AI) { return -AI.Offset; });
 
   const AddrInfo &WithMaxNeeded =
       getMaxOf(MoveInfos, [](const AddrInfo &AI) { return AI.NeedAlign; });

 auto HexagonVectorCombine::getSizeOf(const Value *Val) const -> int {
   return getSizeOf(Val->getType());
 }
 
 auto HexagonVectorCombine::getSizeOf(const Type *Ty) const -> int {
   return DL.getTypeStoreSize(const_cast<Type *>(Ty)).getFixedValue();
 }
 
+auto HexagonVectorCombine::getAllocSizeOf(const Type *Ty) const -> int {
+  return DL.getTypeAllocSize(const_cast<Type *>(Ty)).getFixedValue();
+}
+
 auto HexagonVectorCombine::getTypeAlignment(Type *Ty) const -> int {
   // The actual type may be shorter than the HVX vector, so determine
   // the alignment based on subtarget info.
   if (HST.isTypeForHVX(Ty))
     return HST.getVectorLength();
   return DL.getABITypeAlign(Ty).value();
 }
 
auto HexagonVectorCombine::rescale(IRBuilder<> &Builder, Value *Mask,
   int ToSize = getSizeOf(ToSTy);
   assert(FromSize % ToSize == 0 || ToSize % FromSize == 0);
 
   auto *MaskTy = cast<VectorType>(Mask->getType());
   int FromCount = MaskTy->getElementCount().getFixedValue();
   int ToCount = (FromCount * FromSize) / ToSize;
   assert((FromCount * FromSize) % ToSize == 0);
 
+  auto *FromITy = IntegerType::get(F.getContext(), FromSize * 8);
+  auto *ToITy = IntegerType::get(F.getContext(), ToSize * 8);
+
   // Mask <N x i1> -> sext to <N x FromTy> -> bitcast to <M x ToTy> ->
   // -> trunc to <M x i1>.
   Value *Ext = Builder.CreateSExt(
-      Mask, VectorType::get(FromSTy, FromCount, /*Scalable*/ false));
+      Mask, VectorType::get(FromITy, FromCount, /*Scalable*/ false));
   Value *Cast = Builder.CreateBitCast(
-      Ext, VectorType::get(ToSTy, ToCount, /*Scalable*/ false));
+      Ext, VectorType::get(ToITy, ToCount, /*Scalable*/ false));
   return Builder.CreateTrunc(
       Cast, VectorType::get(getBoolTy(), ToCount, /*Scalable*/ false));
 }
 
 // Bitcast to bytes, and return least significant bits.
 auto HexagonVectorCombine::vlsb(IRBuilder<> &Builder, Value *Val) const
     -> Value * {
   Type *ScalarTy = Val->getType()->getScalarType();
if (!isa<GetElementPtrInst>(Ptr0) || !isa<GetElementPtrInst>(Ptr1))
     return None;
 
   auto *Gep0 = cast<GetElementPtrInst>(Ptr0);
   auto *Gep1 = cast<GetElementPtrInst>(Ptr1);
   if (Gep0->getPointerOperand() != Gep1->getPointerOperand())
     return None;
 
   Builder B(Gep0->getParent());
-  int Scale = DL.getTypeStoreSize(Gep0->getSourceElementType());
+  int Scale = getAllocSizeOf(Gep0->getSourceElementType());
 
   // FIXME: for now only check GEPs with a single index.
   if (Gep0->getNumOperands() != 2 || Gep1->getNumOperands() != 2)
     return None;
 
   Value *Idx0 = Gep0->getOperand(1);
   Value *Idx1 = Gep1->getOperand(1);
 

llvm/test/CodeGen/Hexagon/autohvx/vector-align-addr.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -hexagon-vc -march=hexagon -hexagon-hvx-widen=32 -S < %s | FileCheck %s
+
+; Test that the Hexagon Vector Combine pass computes the address
+; correctly when the loading objects that contain extra padding
+; between successive objects.
+
+target datalayout = "e-m:e-p:32:32:32-a:0-n16:32-i64:64:64-i32:32:32-i16:16:16-i1:8:8-f32:32:32-f64:64:64-v32:32:32-v64:64:64-v512:512:512-v1024:1024:1024-v2048:2048:2048"
+target triple = "hexagon"
+
+define dllexport void @test(i8* %0) local_unnamed_addr #0 {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 0, 3040
+; CHECK-NEXT:    [[TMP2:%.*]] = load i8, i8* undef, align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <19 x i8> undef, i8 [[TMP2]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <19 x i8> [[TMP3]], <19 x i8> undef, <19 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, i8* [[TMP0:%.*]], i32 [[TMP1]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to <19 x i8>*
+; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i32 0, 5472
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, i8* [[TMP0]], i32 [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i8* [[TMP8]] to <19 x i8>*
+; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint <19 x i8>* [[TMP6]] to i32
+; CHECK-NEXT:    [[TMP11:%.*]] = and i32 [[TMP10]], -128
+; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP11]] to <19 x i8>*
+; CHECK-NEXT:    [[TMP13:%.*]] = ptrtoint <19 x i8>* [[TMP6]] to i32
+; CHECK-NEXT:    [[TMP14:%.*]] = bitcast <19 x i8>* [[TMP12]] to i8*
+; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i8, i8* [[TMP14]], i32 0
+; CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP15]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP17:%.*]] = load <128 x i8>, <128 x i8>* [[TMP16]], align 128
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 4
+; CHECK-NEXT:    [[TMP19:%.*]] = bitcast <19 x i8>* [[TMP18]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP20:%.*]] = load <128 x i8>, <128 x i8>* [[TMP19]], align 128
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 8
+; CHECK-NEXT:    [[TMP22:%.*]] = bitcast <19 x i8>* [[TMP21]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP23:%.*]] = load <128 x i8>, <128 x i8>* [[TMP22]], align 128
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 12
+; CHECK-NEXT:    [[TMP25:%.*]] = bitcast <19 x i8>* [[TMP24]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP26:%.*]] = load <128 x i8>, <128 x i8>* [[TMP25]], align 128
+; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 16
+; CHECK-NEXT:    [[TMP28:%.*]] = bitcast <19 x i8>* [[TMP27]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP29:%.*]] = load <128 x i8>, <128 x i8>* [[TMP28]], align 128
+; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 20
+; CHECK-NEXT:    [[TMP31:%.*]] = bitcast <19 x i8>* [[TMP30]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP32:%.*]] = load <128 x i8>, <128 x i8>* [[TMP31]], align 128
+; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 24
+; CHECK-NEXT:    [[TMP34:%.*]] = bitcast <19 x i8>* [[TMP33]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP35:%.*]] = load <128 x i8>, <128 x i8>* [[TMP34]], align 128
+; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 28
+; CHECK-NEXT:    [[TMP37:%.*]] = bitcast <19 x i8>* [[TMP36]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP38:%.*]] = load <128 x i8>, <128 x i8>* [[TMP37]], align 128
+; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 32
+; CHECK-NEXT:    [[TMP40:%.*]] = bitcast <19 x i8>* [[TMP39]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP41:%.*]] = load <128 x i8>, <128 x i8>* [[TMP40]], align 128
+; CHECK-NEXT:    [[TMP42:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 36
+; CHECK-NEXT:    [[TMP43:%.*]] = bitcast <19 x i8>* [[TMP42]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP44:%.*]] = load <128 x i8>, <128 x i8>* [[TMP43]], align 128
+; CHECK-NEXT:    [[TMP45:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 40
+; CHECK-NEXT:    [[TMP46:%.*]] = bitcast <19 x i8>* [[TMP45]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP47:%.*]] = load <128 x i8>, <128 x i8>* [[TMP46]], align 128
+; CHECK-NEXT:    [[TMP48:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 44
+; CHECK-NEXT:    [[TMP49:%.*]] = bitcast <19 x i8>* [[TMP48]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP50:%.*]] = load <128 x i8>, <128 x i8>* [[TMP49]], align 128
+; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 48
+; CHECK-NEXT:    [[TMP52:%.*]] = bitcast <19 x i8>* [[TMP51]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP53:%.*]] = load <128 x i8>, <128 x i8>* [[TMP52]], align 128
+; CHECK-NEXT:    [[TMP54:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 52
+; CHECK-NEXT:    [[TMP55:%.*]] = bitcast <19 x i8>* [[TMP54]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP56:%.*]] = load <128 x i8>, <128 x i8>* [[TMP55]], align 128
+; CHECK-NEXT:    [[TMP57:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 56
+; CHECK-NEXT:    [[TMP58:%.*]] = bitcast <19 x i8>* [[TMP57]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP59:%.*]] = load <128 x i8>, <128 x i8>* [[TMP58]], align 128
+; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 60
+; CHECK-NEXT:    [[TMP61:%.*]] = bitcast <19 x i8>* [[TMP60]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP62:%.*]] = load <128 x i8>, <128 x i8>* [[TMP61]], align 128
+; CHECK-NEXT:    [[TMP63:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 64
+; CHECK-NEXT:    [[TMP64:%.*]] = bitcast <19 x i8>* [[TMP63]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP65:%.*]] = load <128 x i8>, <128 x i8>* [[TMP64]], align 128
+; CHECK-NEXT:    [[TMP66:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 68
+; CHECK-NEXT:    [[TMP67:%.*]] = bitcast <19 x i8>* [[TMP66]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP68:%.*]] = load <128 x i8>, <128 x i8>* [[TMP67]], align 128
+; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 72
+; CHECK-NEXT:    [[TMP70:%.*]] = bitcast <19 x i8>* [[TMP69]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP71:%.*]] = load <128 x i8>, <128 x i8>* [[TMP70]], align 128
+; CHECK-NEXT:    [[TMP72:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 76
+; CHECK-NEXT:    [[TMP73:%.*]] = bitcast <19 x i8>* [[TMP72]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP74:%.*]] = load <128 x i8>, <128 x i8>* [[TMP73]], align 128
+; CHECK-NEXT:    [[TMP75:%.*]] = getelementptr <19 x i8>, <19 x i8>* [[TMP12]], i32 80
+; CHECK-NEXT:    [[TMP76:%.*]] = bitcast <19 x i8>* [[TMP75]] to <128 x i8>*
+; CHECK-NEXT:    [[TMP77:%.*]] = load <128 x i8>, <128 x i8>* [[TMP76]], align 128
+; CHECK-NEXT:    [[TMP78:%.*]] = bitcast <128 x i8> [[TMP20]] to <32 x i32>
+; CHECK-NEXT:    [[TMP79:%.*]] = bitcast <128 x i8> [[TMP17]] to <32 x i32>
+; CHECK-NEXT:    [[TMP80:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP78]], <32 x i32> [[TMP79]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP81:%.*]] = bitcast <32 x i32> [[TMP80]] to <128 x i8>
+; CHECK-NEXT:    [[TMP82:%.*]] = bitcast <128 x i8> [[TMP23]] to <32 x i32>
+; CHECK-NEXT:    [[TMP83:%.*]] = bitcast <128 x i8> [[TMP20]] to <32 x i32>
+; CHECK-NEXT:    [[TMP84:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP82]], <32 x i32> [[TMP83]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP85:%.*]] = bitcast <32 x i32> [[TMP84]] to <128 x i8>
+; CHECK-NEXT:    [[TMP86:%.*]] = bitcast <128 x i8> [[TMP26]] to <32 x i32>
+; CHECK-NEXT:    [[TMP87:%.*]] = bitcast <128 x i8> [[TMP23]] to <32 x i32>
+; CHECK-NEXT:    [[TMP88:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP86]], <32 x i32> [[TMP87]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP89:%.*]] = bitcast <32 x i32> [[TMP88]] to <128 x i8>
+; CHECK-NEXT:    [[TMP90:%.*]] = bitcast <128 x i8> [[TMP29]] to <32 x i32>
+; CHECK-NEXT:    [[TMP91:%.*]] = bitcast <128 x i8> [[TMP26]] to <32 x i32>
+; CHECK-NEXT:    [[TMP92:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP90]], <32 x i32> [[TMP91]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP93:%.*]] = bitcast <32 x i32> [[TMP92]] to <128 x i8>
+; CHECK-NEXT:    [[TMP94:%.*]] = bitcast <128 x i8> [[TMP32]] to <32 x i32>
+; CHECK-NEXT:    [[TMP95:%.*]] = bitcast <128 x i8> [[TMP29]] to <32 x i32>
+; CHECK-NEXT:    [[TMP96:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP94]], <32 x i32> [[TMP95]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP97:%.*]] = bitcast <32 x i32> [[TMP96]] to <128 x i8>
+; CHECK-NEXT:    [[TMP98:%.*]] = bitcast <128 x i8> [[TMP35]] to <32 x i32>
+; CHECK-NEXT:    [[TMP99:%.*]] = bitcast <128 x i8> [[TMP32]] to <32 x i32>
+; CHECK-NEXT:    [[TMP100:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP98]], <32 x i32> [[TMP99]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP101:%.*]] = bitcast <32 x i32> [[TMP100]] to <128 x i8>
+; CHECK-NEXT:    [[TMP102:%.*]] = bitcast <128 x i8> [[TMP38]] to <32 x i32>
+; CHECK-NEXT:    [[TMP103:%.*]] = bitcast <128 x i8> [[TMP35]] to <32 x i32>
+; CHECK-NEXT:    [[TMP104:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP102]], <32 x i32> [[TMP103]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP105:%.*]] = bitcast <32 x i32> [[TMP104]] to <128 x i8>
+; CHECK-NEXT:    [[TMP106:%.*]] = bitcast <128 x i8> [[TMP41]] to <32 x i32>
+; CHECK-NEXT:    [[TMP107:%.*]] = bitcast <128 x i8> [[TMP38]] to <32 x i32>
+; CHECK-NEXT:    [[TMP108:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP106]], <32 x i32> [[TMP107]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP109:%.*]] = bitcast <32 x i32> [[TMP108]] to <128 x i8>
+; CHECK-NEXT:    [[TMP110:%.*]] = bitcast <128 x i8> [[TMP44]] to <32 x i32>
+; CHECK-NEXT:    [[TMP111:%.*]] = bitcast <128 x i8> [[TMP41]] to <32 x i32>
+; CHECK-NEXT:    [[TMP112:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP110]], <32 x i32> [[TMP111]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP113:%.*]] = bitcast <32 x i32> [[TMP112]] to <128 x i8>
+; CHECK-NEXT:    [[TMP114:%.*]] = bitcast <128 x i8> [[TMP47]] to <32 x i32>
+; CHECK-NEXT:    [[TMP115:%.*]] = bitcast <128 x i8> [[TMP44]] to <32 x i32>
+; CHECK-NEXT:    [[TMP116:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP114]], <32 x i32> [[TMP115]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP117:%.*]] = bitcast <32 x i32> [[TMP116]] to <128 x i8>
+; CHECK-NEXT:    [[TMP118:%.*]] = bitcast <128 x i8> [[TMP50]] to <32 x i32>
+; CHECK-NEXT:    [[TMP119:%.*]] = bitcast <128 x i8> [[TMP47]] to <32 x i32>
+; CHECK-NEXT:    [[TMP120:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP118]], <32 x i32> [[TMP119]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP121:%.*]] = bitcast <32 x i32> [[TMP120]] to <128 x i8>
+; CHECK-NEXT:    [[TMP122:%.*]] = bitcast <128 x i8> [[TMP53]] to <32 x i32>
+; CHECK-NEXT:    [[TMP123:%.*]] = bitcast <128 x i8> [[TMP50]] to <32 x i32>
+; CHECK-NEXT:    [[TMP124:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP122]], <32 x i32> [[TMP123]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP125:%.*]] = bitcast <32 x i32> [[TMP124]] to <128 x i8>
+; CHECK-NEXT:    [[TMP126:%.*]] = bitcast <128 x i8> [[TMP56]] to <32 x i32>
+; CHECK-NEXT:    [[TMP127:%.*]] = bitcast <128 x i8> [[TMP53]] to <32 x i32>
+; CHECK-NEXT:    [[TMP128:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP126]], <32 x i32> [[TMP127]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP129:%.*]] = bitcast <32 x i32> [[TMP128]] to <128 x i8>
+; CHECK-NEXT:    [[TMP130:%.*]] = bitcast <128 x i8> [[TMP59]] to <32 x i32>
+; CHECK-NEXT:    [[TMP131:%.*]] = bitcast <128 x i8> [[TMP56]] to <32 x i32>
+; CHECK-NEXT:    [[TMP132:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP130]], <32 x i32> [[TMP131]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP133:%.*]] = bitcast <32 x i32> [[TMP132]] to <128 x i8>
+; CHECK-NEXT:    [[TMP134:%.*]] = bitcast <128 x i8> [[TMP62]] to <32 x i32>
+; CHECK-NEXT:    [[TMP135:%.*]] = bitcast <128 x i8> [[TMP59]] to <32 x i32>
+; CHECK-NEXT:    [[TMP136:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP134]], <32 x i32> [[TMP135]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP137:%.*]] = bitcast <32 x i32> [[TMP136]] to <128 x i8>
+; CHECK-NEXT:    [[TMP138:%.*]] = bitcast <128 x i8> [[TMP65]] to <32 x i32>
+; CHECK-NEXT:    [[TMP139:%.*]] = bitcast <128 x i8> [[TMP62]] to <32 x i32>
+; CHECK-NEXT:    [[TMP140:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP138]], <32 x i32> [[TMP139]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP141:%.*]] = bitcast <32 x i32> [[TMP140]] to <128 x i8>
+; CHECK-NEXT:    [[TMP142:%.*]] = bitcast <128 x i8> [[TMP68]] to <32 x i32>
+; CHECK-NEXT:    [[TMP143:%.*]] = bitcast <128 x i8> [[TMP65]] to <32 x i32>
+; CHECK-NEXT:    [[TMP144:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP142]], <32 x i32> [[TMP143]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP145:%.*]] = bitcast <32 x i32> [[TMP144]] to <128 x i8>
+; CHECK-NEXT:    [[TMP146:%.*]] = bitcast <128 x i8> [[TMP71]] to <32 x i32>
+; CHECK-NEXT:    [[TMP147:%.*]] = bitcast <128 x i8> [[TMP68]] to <32 x i32>
+; CHECK-NEXT:    [[TMP148:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP146]], <32 x i32> [[TMP147]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP149:%.*]] = bitcast <32 x i32> [[TMP148]] to <128 x i8>
+; CHECK-NEXT:    [[TMP150:%.*]] = bitcast <128 x i8> [[TMP74]] to <32 x i32>
+; CHECK-NEXT:    [[TMP151:%.*]] = bitcast <128 x i8> [[TMP71]] to <32 x i32>
+; CHECK-NEXT:    [[TMP152:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP150]], <32 x i32> [[TMP151]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP153:%.*]] = bitcast <32 x i32> [[TMP152]] to <128 x i8>
+; CHECK-NEXT:    [[TMP154:%.*]] = bitcast <128 x i8> [[TMP77]] to <32 x i32>
+; CHECK-NEXT:    [[TMP155:%.*]] = bitcast <128 x i8> [[TMP74]] to <32 x i32>
+; CHECK-NEXT:    [[TMP156:%.*]] = call <32 x i32> @llvm.hexagon.V6.valignb.128B(<32 x i32> [[TMP154]], <32 x i32> [[TMP155]], i32 [[TMP13]])
+; CHECK-NEXT:    [[TMP157:%.*]] = bitcast <32 x i32> [[TMP156]] to <128 x i8>
+; CHECK-NEXT:    [[TMP158:%.*]] = shufflevector <128 x i8> [[TMP81]], <128 x i8> undef, <256 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 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>
+; CHECK-NEXT:    [[TMP159:%.*]] = shufflevector <256 x i8> undef, <256 x i8> [[TMP158]], <256 x i32> <i32 256, i32 257, i32 258, i32 259, i32 260, i32 261, i32 262, i32 263, i32 264, i32 265, i32 266, i32 267, i32 268, i32 269, i32 270, i32 271, i32 272, i32 273, i32 274, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 129, i32 130, i32 131, i32 132, i32 133, i32 134, i32 135, i32 136, i32 137, i32 138, i32 139, i32 140, i32 141, i32 142, i32 143, i32 144, i32 145, i32 146, i32 147, i32 148, i32 149, i32 150, i32 151, i32 152, i32 153, i32 154, i32 155, i32 156, i32 157, i32 158, i32 159, i32 160, i32 161, i32 162, i32 163, i32 164, i32 165, i32 166, i32 167, i32 168, i32 169, i32 170, i32 171, i32 172, i32 173, i32 174, i32 175, i32 176, i32 177, i32 178, i32 179, i32 180, i32 181, i32 182, i32 183, i32 184, i32 185, i32 186, i32 187, i32 188, i32 189, i32 190, i32 191, i32 192, i32 193, i32 194, i32 195, i32 196, i32 197, i32 198, i32 199, i32 200, i32 201, i32 202, i32 203, i32 204, i32 205, i32 206, i32 207, i32 208, i32 209, i32 210, i32 211, i32 212, i32 213, i32 214, i32 215, i32 216, i32 217, i32 218, i32 219, i32 220, i32 221, i32 222, i32 223, i32 224, i32 225, i32 226, i32 227, i32 228, i32 229, i32 230, i32 231, i32 232, i32 233, i32 234, i32 235, i32 236, i32 237, i32 238, i32 239, i32 240, i32 241, i32 242, i32 243, i32 244, i32 245, i32 246, i32 247, i32 248, i32 249, i32 250, i32 251, i32 252, i32 253, i32 254, i32 255>
+; CHECK-NEXT:    [[TMP160:%.*]] = shufflevector <256 x i8> [[TMP159]], <256 x i8> [[TMP159]], <19 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 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+; CHECK-NEXT:    [[TMP161:%.*]] = select <19 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <19 x i8> [[TMP160]], <19 x i8> undef
+; CHECK-NEXT:    [[TMP162:%.*]] = shufflevector <128 x i8> [[TMP157]], <128 x i8> undef, <256 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 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128, i32 128>
+; CHECK-NEXT:    [[TMP163:%.*]] = shufflevector <256 x i8> undef, <256 x i8> [[TMP162]], <256 x i32> <i32 256, i32 257, i32 258, i32 259, i32 260, i32 261, i32 262, i32 263, i32 264, i32 265, i32 266, i32 267, i32 268, i32 269, i32 270, i32 271, i32 272, i32 273, i32 274, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 129, i32 130, i32 131, i32 132, i32 133, i32 134, i32 135, i32 136, i32 137, i32 138, i32 139, i32 140, i32 141, i32 142, i32 143, i32 144, i32 145, i32 146, i32 147, i32 148, i32 149, i32 150, i32 151, i32 152, i32 153, i32 154, i32 155, i32 156, i32 157, i32 158, i32 159, i32 160, i32 161, i32 162, i32 163, i32 164, i32 165, i32 166, i32 167, i32 168, i32 169, i32 170, i32 171, i32 172, i32 173, i32 174, i32 175, i32 176, i32 177, i32 178, i32 179, i32 180, i32 181, i32 182, i32 183, i32 184, i32 185, i32 186, i32 187, i32 188, i32 189, i32 190, i32 191, i32 192, i32 193, i32 194, i32 195, i32 196, i32 197, i32 198, i32 199, i32 200, i32 201, i32 202, i32 203, i32 204, i32 205, i32 206, i32 207, i32 208, i32 209, i32 210, i32 211, i32 212, i32 213, i32 214, i32 215, i32 216, i32 217, i32 218, i32 219, i32 220, i32 221, i32 222, i32 223, i32 224, i32 225, i32 226, i32 227, i32 228, i32 229, i32 230, i32 231, i32 232, i32 233, i32 234, i32 235, i32 236, i32 237, i32 238, i32 239, i32 240, i32 241, i32 242, i32 243, i32 244, i32 245, i32 246, i32 247, i32 248, i32 249, i32 250, i32 251, i32 252, i32 253, i32 254, i32 255>
+; CHECK-NEXT:    [[TMP164:%.*]] = shufflevector <256 x i8> [[TMP163]], <256 x i8> [[TMP163]], <19 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 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18>
+; CHECK-NEXT:    [[TMP165:%.*]] = select <19 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <19 x i8> [[TMP164]], <19 x i8> undef
+; CHECK-NEXT:    [[TMP166:%.*]] = mul <19 x i8> [[TMP4]], [[TMP161]]
+; CHECK-NEXT:    [[TMP167:%.*]] = add <19 x i8> [[TMP166]], zeroinitializer
+; CHECK-NEXT:    [[TMP168:%.*]] = add <19 x i8> zeroinitializer, [[TMP167]]
+; CHECK-NEXT:    [[TMP169:%.*]] = add <19 x i8> zeroinitializer, [[TMP168]]
+; CHECK-NEXT:    [[TMP170:%.*]] = add <19 x i8> zeroinitializer, [[TMP169]]
+; CHECK-NEXT:    [[TMP171:%.*]] = add <19 x i8> zeroinitializer, [[TMP170]]
+; CHECK-NEXT:    [[TMP172:%.*]] = add <19 x i8> zeroinitializer, [[TMP171]]
+; CHECK-NEXT:    [[TMP173:%.*]] = add <19 x i8> zeroinitializer, [[TMP172]]
+; CHECK-NEXT:    [[TMP174:%.*]] = add <19 x i8> zeroinitializer, [[TMP173]]
+; CHECK-NEXT:    [[TMP175:%.*]] = add <19 x i8> zeroinitializer, [[TMP174]]
+; CHECK-NEXT:    [[TMP176:%.*]] = add <19 x i8> zeroinitializer, [[TMP175]]
+; CHECK-NEXT:    [[TMP177:%.*]] = add <19 x i8> zeroinitializer, [[TMP176]]
+; CHECK-NEXT:    [[TMP178:%.*]] = load i8, i8* undef, align 1
+; CHECK-NEXT:    [[TMP179:%.*]] = insertelement <19 x i8> undef, i8 [[TMP178]], i32 0
+; CHECK-NEXT:    [[TMP180:%.*]] = shufflevector <19 x i8> [[TMP179]], <19 x i8> undef, <19 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP181:%.*]] = mul <19 x i8> [[TMP180]], [[TMP165]]
+; CHECK-NEXT:    [[TMP182:%.*]] = add <19 x i8> [[TMP181]], [[TMP177]]
+; CHECK-NEXT:    [[TMP183:%.*]] = add <19 x i8> zeroinitializer, [[TMP182]]
+; CHECK-NEXT:    [[TMP184:%.*]] = add <19 x i8> zeroinitializer, [[TMP183]]
+; CHECK-NEXT:    [[TMP185:%.*]] = add <19 x i8> zeroinitializer, [[TMP184]]
+; CHECK-NEXT:    [[TMP186:%.*]] = bitcast i8* [[TMP0]] to <19 x i8>*
+; CHECK-NEXT:    store <19 x i8> [[TMP185]], <19 x i8>* [[TMP186]], align 1
+; CHECK-NEXT:    ret void
+;
+entry:
+  %1 = add nuw nsw i32 0, 3040
+  %2 = load i8, i8* undef, align 1
+  %3 = insertelement <19 x i8> undef, i8 %2, i32 0
+  %4 = shufflevector <19 x i8> %3, <19 x i8> undef, <19 x i32> zeroinitializer
+  %5 = getelementptr inbounds i8, i8* %0, i32 %1
+  %6 = bitcast i8* %5 to <19 x i8>*
+  %7 = load <19 x i8>, <19 x i8>* %6, align 1
+  %8 = mul <19 x i8> %4, %7
+  %9 = add <19 x i8> %8, zeroinitializer
+  %10 = add <19 x i8> zeroinitializer, %9
+  %11 = add <19 x i8> zeroinitializer, %10
+  %12 = add <19 x i8> zeroinitializer, %11
+  %13 = add <19 x i8> zeroinitializer, %12
+  %14 = add <19 x i8> zeroinitializer, %13
+  %15 = add <19 x i8> zeroinitializer, %14
+  %16 = add <19 x i8> zeroinitializer, %15
+  %17 = add <19 x i8> zeroinitializer, %16
+  %18 = add <19 x i8> zeroinitializer, %17
+  %19 = add <19 x i8> zeroinitializer, %18
+  %20 = load i8, i8* undef, align 1
+  %21 = insertelement <19 x i8> undef, i8 %20, i32 0
+  %22 = shufflevector <19 x i8> %21, <19 x i8> undef, <19 x i32> zeroinitializer
+  %23 = add nuw nsw i32 0, 5472
+  %24 = getelementptr inbounds i8, i8* %0, i32 %23
+  %25 = bitcast i8* %24 to <19 x i8>*
+  %26 = load <19 x i8>, <19 x i8>* %25, align 1
+  %27 = mul <19 x i8> %22, %26
+  %28 = add <19 x i8> %27, %19
+  %29 = add <19 x i8> zeroinitializer, %28
+  %30 = add <19 x i8> zeroinitializer, %29
+  %31 = add <19 x i8> zeroinitializer, %30
+  %32 = bitcast i8* %0 to <19 x i8>*
+  store <19 x i8> %31, <19 x i8>* %32, align 1
+  ret void
+}
+
+attributes #0 = { "target-features"="+hvxv66,+hvx-length128b" }