[ARM] MVE interleaving load and stores.

[dmgreen]

Summary

Now that we have the intrinsics, we can add VLD2/4 and VST2/4 lowering for MVE. This works the same way as Neon, recognising the load/shuffles combination and converting them into intrinsics in a pre-isel pass, which just calls getMaxSupportedInterleaveFactor, lowerInterleavedLoad and lowerInterleavedStore.

The main difference to Neon is that we do not have a VLD3 instruction. Otherwise most of the code works very similarly, with just some minor differences in the form of the intrinsics to work around. VLD3 is disabled by making isLegalInterleavedAccessType return false for those cases.

We may need some other future adjustments, such as VLD4 take up half the available registers so should maybe cost more. This patch should get the basics in though.

[samparker]

llvm/lib/Target/ARM/ARMISelLowering.cpp

16984

What is Factor here? Is it not the number of elements..?

17040

debug

17212

How about performing this with a loop?

[dmgreen]

llvm/lib/Target/ARM/ARMISelLowering.cpp

16984

Factor is the "n" in "vldn". So 2 or 4 for MVE, but also 3 for Neon. It comes from the elements of the shuffle, like the example in the comment at the start of this function.

We really just need it to say that a VLD3 isn't legal on MVE.

17040

Ah!

17212

Yeah. Nice suggestion. It involves a pop because of the last item changing, but looks better.

[dmgreen]

Thanks. I've updated with the suggestions.

Also:

• Added a load of extra tests.
• Updated the cost model a little, notably for smaller than legal types that look like interleaves but actually use a single load with a vrev or a vmovn or something like it.
• Err, formatted =)

[samparker]

llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp

817

if this is handling the smaller vector types, shouldn't this be <= 64 instead? Is the 2 below 'Factor'? If so it would be good to use the variable name or a clear explanation of what it is.

[dmgreen]

llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp

817

Oh, yeah, it's SrcVecTy, so yes <= 64 will be fine (128 will already be handled in the above if).

The 2 below is VLDR, VREV or VLDR; VMOVN. 2 Instructions though. I'll add a comment.

[dmgreen]

Updated "vmovn" cost check.

[samparker]

Nice one. LGTM

llvm/lib/Target/ARM/ARMISelLowering.h

public:
     bool shouldExpandShift(SelectionDAG &DAG, SDNode *N) const override;
 
     CCAssignFn *CCAssignFnForCall(CallingConv::ID CC, bool isVarArg) const;
     CCAssignFn *CCAssignFnForReturn(CallingConv::ID CC, bool isVarArg) const;
 
     /// Returns true if \p VecTy is a legal interleaved access type. This
     /// function checks the vector element type and the overall width of the
     /// vector.
-    bool isLegalInterleavedAccessType(VectorType *VecTy,
+    bool isLegalInterleavedAccessType(unsigned Factor, VectorType *VecTy,
                                       const DataLayout &DL) const;
 
     bool alignLoopsWithOptSize() const override;
 
     /// Returns the number of interleaved accesses that will be generated when
     /// lowering accesses of the given type.
     unsigned getNumInterleavedAccesses(VectorType *VecTy,
                                        const DataLayout &DL) const;

llvm/lib/Target/ARM/ARMISelLowering.cpp

 /// will generate when lowering accesses of the given type.
 unsigned
 ARMTargetLowering::getNumInterleavedAccesses(VectorType *VecTy,
                                              const DataLayout &DL) const {
   return (DL.getTypeSizeInBits(VecTy) + 127) / 128;
 }
 
 bool ARMTargetLowering::isLegalInterleavedAccessType(
-    VectorType *VecTy, const DataLayout &DL) const {
+    unsigned Factor, VectorType *VecTy, const DataLayout &DL) const {
 
   unsigned VecSize = DL.getTypeSizeInBits(VecTy);
   unsigned ElSize = DL.getTypeSizeInBits(VecTy->getElementType());
 
+  if (!Subtarget->hasNEON() && !Subtarget->hasMVEIntegerOps())
+    return false;
+
   // Ensure the vector doesn't have f16 elements. Even though we could do an
   // i16 vldN, we can't hold the f16 vectors and will end up converting via
   // f32.
-  if (VecTy->getElementType()->isHalfTy())
+  if (Subtarget->hasNEON() && VecTy->getElementType()->isHalfTy())
+    return false;
+  if (Subtarget->hasMVEIntegerOps() && Factor == 3)
     return false;
 
   // Ensure the number of vector elements is greater than 1.
   if (VecTy->getNumElements() < 2)
     return false;
 
   // Ensure the element type is legal.
   if (ElSize != 8 && ElSize != 16 && ElSize != 32)
     return false;
 
   // Ensure the total vector size is 64 or a multiple of 128. Types larger than
   // 128 will be split into multiple interleaved accesses.
-  return VecSize == 64 || VecSize % 128 == 0;
+  if (Subtarget->hasNEON() && VecSize == 64)
+    return true;
+  return VecSize % 128 == 0;
 }
 
 unsigned ARMTargetLowering::getMaxSupportedInterleaveFactor() const {
   if (Subtarget->hasNEON())
     return 4;
+  if (Subtarget->hasMVEIntegerOps())
+    return 4;
   return TargetLoweringBase::getMaxSupportedInterleaveFactor();
 }
 
 /// Lower an interleaved load into a vldN intrinsic.
 ///
 /// E.g. Lower an interleaved load (Factor = 2):
 ///        %wide.vec = load <8 x i32>, <8 x i32>* %ptr, align 4
 ///        %v0 = shuffle %wide.vec, undef, <0, 2, 4, 6>  ; Extract even elements
bool ARMTargetLowering::lowerInterleavedLoad(
   VectorType *VecTy = Shuffles[0]->getType();
   Type *EltTy = VecTy->getVectorElementType();
 
   const DataLayout &DL = LI->getModule()->getDataLayout();
 
   // Skip if we do not have NEON and skip illegal vector types. We can
   // "legalize" wide vector types into multiple interleaved accesses as long as
   // the vector types are divisible by 128.
-  if (!Subtarget->hasNEON() || !isLegalInterleavedAccessType(VecTy, DL))
+  if (!isLegalInterleavedAccessType(Factor, VecTy, DL))
     return false;
 
   unsigned NumLoads = getNumInterleavedAccesses(VecTy, DL);
 
   // A pointer vector can not be the return type of the ldN intrinsics. Need to
   // load integer vectors first and then convert to pointer vectors.
   if (EltTy->isPointerTy())
     VecTy =
if (NumLoads > 1) {
     // element type.
     BaseAddr = Builder.CreateBitCast(
         BaseAddr, VecTy->getVectorElementType()->getPointerTo(
                       LI->getPointerAddressSpace()));
   }
 
   assert(isTypeLegal(EVT::getEVT(VecTy)) && "Illegal vldN vector type!");
 
+  auto createLoadIntrinsic = [&](Value *BaseAddr) {
+    if (Subtarget->hasNEON()) {
       Type *Int8Ptr = Builder.getInt8PtrTy(LI->getPointerAddressSpace());
       Type *Tys[] = {VecTy, Int8Ptr};
       static const Intrinsic::ID LoadInts[3] = {Intrinsic::arm_neon_vld2,
                                                 Intrinsic::arm_neon_vld3,
                                                 Intrinsic::arm_neon_vld4};
       Function *VldnFunc =
           Intrinsic::getDeclaration(LI->getModule(), LoadInts[Factor - 2], Tys);
 
+      SmallVector<Value *, 2> Ops;
+      Ops.push_back(Builder.CreateBitCast(BaseAddr, Int8Ptr));
+      Ops.push_back(Builder.getInt32(LI->getAlignment()));
+
+      return Builder.CreateCall(VldnFunc, Ops, "vldN");
+    } else {
+      assert((Factor == 2 || Factor == 4) &&
+             "expected interleave factor of 2 or 4 for MVE");
+      Intrinsic::ID LoadInts =
+          Factor == 2 ? Intrinsic::arm_mve_vld2q : Intrinsic::arm_mve_vld4q;
+      Type *VecEltTy = VecTy->getVectorElementType()->getPointerTo(
+          LI->getPointerAddressSpace());
+      Type *Tys[] = {VecTy, VecEltTy};
+      Function *VldnFunc =
+          Intrinsic::getDeclaration(LI->getModule(), LoadInts, Tys);
+
+      SmallVector<Value *, 2> Ops;
+      Ops.push_back(Builder.CreateBitCast(BaseAddr, VecEltTy));
+      return Builder.CreateCall(VldnFunc, Ops, "vldN");
+    }
+  };
+
   // Holds sub-vectors extracted from the load intrinsic return values. The
   // sub-vectors are associated with the shufflevector instructions they will
   // replace.
   DenseMap<ShuffleVectorInst *, SmallVector<Value *, 4>> SubVecs;
 
   for (unsigned LoadCount = 0; LoadCount < NumLoads; ++LoadCount) {
     // If we're generating more than one load, compute the base address of
     // subsequent loads as an offset from the previous.
     if (LoadCount > 0)
       BaseAddr =
           Builder.CreateConstGEP1_32(VecTy->getVectorElementType(), BaseAddr,
                                      VecTy->getVectorNumElements() * Factor);
 
-    SmallVector<Value *, 2> Ops;
-    Ops.push_back(Builder.CreateBitCast(BaseAddr, Int8Ptr));
-    Ops.push_back(Builder.getInt32(LI->getAlignment()));
-
-    CallInst *VldN = Builder.CreateCall(VldnFunc, Ops, "vldN");
+    CallInst *VldN = createLoadIntrinsic(BaseAddr);
 
     // Replace uses of each shufflevector with the corresponding vector loaded
     // by ldN.
     for (unsigned i = 0; i < Shuffles.size(); i++) {
       ShuffleVectorInst *SV = Shuffles[i];
       unsigned Index = Indices[i];
 
       Value *SubVec = Builder.CreateExtractValue(VldN, Index);
bool ARMTargetLowering::lowerInterleavedStore(StoreInst *SI,
   Type *EltTy = VecTy->getVectorElementType();
   VectorType *SubVecTy = VectorType::get(EltTy, LaneLen);
 
   const DataLayout &DL = SI->getModule()->getDataLayout();
 
   // Skip if we do not have NEON and skip illegal vector types. We can
   // "legalize" wide vector types into multiple interleaved accesses as long as
   // the vector types are divisible by 128.
-  if (!Subtarget->hasNEON() || !isLegalInterleavedAccessType(SubVecTy, DL))
+  if (!isLegalInterleavedAccessType(Factor, SubVecTy, DL))
     return false;
 
   unsigned NumStores = getNumInterleavedAccesses(SubVecTy, DL);
 
   Value *Op0 = SVI->getOperand(0);
   Value *Op1 = SVI->getOperand(1);
   IRBuilder<> Builder(SI);
 
BaseAddr = Builder.CreateBitCast(
         BaseAddr, SubVecTy->getVectorElementType()->getPointerTo(
                       SI->getPointerAddressSpace()));
   }
 
   assert(isTypeLegal(EVT::getEVT(SubVecTy)) && "Illegal vstN vector type!");
 
   auto Mask = SVI->getShuffleMask();
 
-  Type *Int8Ptr = Builder.getInt8PtrTy(SI->getPointerAddressSpace());
-  Type *Tys[] = {Int8Ptr, SubVecTy};
+  auto createStoreIntrinsic = [&](Value *BaseAddr,
+                                  SmallVectorImpl<Value *> &Shuffles) {
+    if (Subtarget->hasNEON()) {
       static const Intrinsic::ID StoreInts[3] = {Intrinsic::arm_neon_vst2,
                                                  Intrinsic::arm_neon_vst3,
                                                  Intrinsic::arm_neon_vst4};
+      Type *Int8Ptr = Builder.getInt8PtrTy(SI->getPointerAddressSpace());
+      Type *Tys[] = {Int8Ptr, SubVecTy};
+
+      Function *VstNFunc = Intrinsic::getDeclaration(
+          SI->getModule(), StoreInts[Factor - 2], Tys);
+
+      SmallVector<Value *, 6> Ops;
+      Ops.push_back(Builder.CreateBitCast(BaseAddr, Int8Ptr));
+      for (auto S : Shuffles)
+        Ops.push_back(S);
+      Ops.push_back(Builder.getInt32(SI->getAlignment()));
+      Builder.CreateCall(VstNFunc, Ops);
+    } else {
+      assert((Factor == 2 || Factor == 4) &&
+             "expected interleave factor of 2 or 4 for MVE");
+      Intrinsic::ID StoreInts =
+          Factor == 2 ? Intrinsic::arm_mve_vst2q : Intrinsic::arm_mve_vst4q;
+      Type *EltPtrTy = SubVecTy->getVectorElementType()->getPointerTo(
+          SI->getPointerAddressSpace());
+      Type *Tys[] = {EltPtrTy, SubVecTy};
+      Function *VstNFunc =
+          Intrinsic::getDeclaration(SI->getModule(), StoreInts, Tys);
+
+      SmallVector<Value *, 6> Ops;
+      Ops.push_back(Builder.CreateBitCast(BaseAddr, EltPtrTy));
+      for (auto S : Shuffles)
+        Ops.push_back(S);
+      for (unsigned F = 0; F < Factor; F++) {
+        Ops.push_back(Builder.getInt32(F));
+        Builder.CreateCall(VstNFunc, Ops);
+        Ops.pop_back();
+      }
+    }
+  };
 
   for (unsigned StoreCount = 0; StoreCount < NumStores; ++StoreCount) {
     // If we generating more than one store, we compute the base address of
     // subsequent stores as an offset from the previous.
     if (StoreCount > 0)
       BaseAddr = Builder.CreateConstGEP1_32(SubVecTy->getVectorElementType(),
                                             BaseAddr, LaneLen * Factor);
 
-    SmallVector<Value *, 6> Ops;
-    Ops.push_back(Builder.CreateBitCast(BaseAddr, Int8Ptr));
-
-    Function *VstNFunc =
-        Intrinsic::getDeclaration(SI->getModule(), StoreInts[Factor - 2], Tys);
+    SmallVector<Value *, 4> Shuffles;
 
     // Split the shufflevector operands into sub vectors for the new vstN call.
     for (unsigned i = 0; i < Factor; i++) {
       unsigned IdxI = StoreCount * LaneLen * Factor + i;
       if (Mask[IdxI] >= 0) {
-        Ops.push_back(Builder.CreateShuffleVector(
+        Shuffles.push_back(Builder.CreateShuffleVector(
             Op0, Op1, createSequentialMask(Builder, Mask[IdxI], LaneLen, 0)));
       } else {
         unsigned StartMask = 0;
         for (unsigned j = 1; j < LaneLen; j++) {
           unsigned IdxJ = StoreCount * LaneLen * Factor + j;
           if (Mask[IdxJ * Factor + IdxI] >= 0) {
             StartMask = Mask[IdxJ * Factor + IdxI] - IdxJ;
             break;
           }
         }
         // Note: If all elements in a chunk are undefs, StartMask=0!
         // Note: Filling undef gaps with random elements is ok, since
         // those elements were being written anyway (with undefs).
         // In the case of all undefs we're defaulting to using elems from 0
         // Note: StartMask cannot be negative, it's checked in
         // isReInterleaveMask
-        Ops.push_back(Builder.CreateShuffleVector(
+        Shuffles.push_back(Builder.CreateShuffleVector(
             Op0, Op1, createSequentialMask(Builder, StartMask, LaneLen, 0)));
       }
     }
 
-    Ops.push_back(Builder.getInt32(SI->getAlignment()));
-    Builder.CreateCall(VstNFunc, Ops);
+    createStoreIntrinsic(BaseAddr, Shuffles);
   }
   return true;
 }
 
 enum HABaseType {
   HA_UNKNOWN = 0,
   HA_FLOAT,
   HA_DOUBLE,

llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp

if (ST->hasNEON() && Src->isVectorTy() && Alignment != 16 &&
     return LT.first * 4;
   }
   int BaseCost = ST->hasMVEIntegerOps() && Src->isVectorTy()
                      ? ST->getMVEVectorCostFactor()
                      : 1;
   return BaseCost * LT.first;
 }
 
-int ARMTTIImpl::getInterleavedMemoryOpCost(unsigned Opcode, Type *VecTy,
-                                           unsigned Factor,
-                                           ArrayRef<unsigned> Indices,
-                                           unsigned Alignment,
-                                           unsigned AddressSpace,
-                                           bool UseMaskForCond,
-                                           bool UseMaskForGaps) {
+int ARMTTIImpl::getInterleavedMemoryOpCost(
+    unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
+    unsigned Alignment, unsigned AddressSpace, bool UseMaskForCond,
   assert(Factor >= 2 && "Invalid interleave factor");
   assert(isa<VectorType>(VecTy) && "Expect a vector type");
 
   // vldN/vstN doesn't support vector types of i64/f64 element.
   bool EltIs64Bits = DL.getTypeSizeInBits(VecTy->getScalarType()) == 64;
 
   if (Factor <= TLI->getMaxSupportedInterleaveFactor() && !EltIs64Bits &&
       !UseMaskForCond && !UseMaskForGaps) {
     unsigned NumElts = VecTy->getVectorNumElements();
     auto *SubVecTy = VectorType::get(VecTy->getScalarType(), NumElts / Factor);
 
     // vldN/vstN only support legal vector types of size 64 or 128 in bits.
     // Accesses having vector types that are a multiple of 128 bits can be
     // matched to more than one vldN/vstN instruction.
+    int BaseCost = ST->hasMVEIntegerOps() ? ST->getMVEVectorCostFactor() : 1;
     if (NumElts % Factor == 0 &&
-        TLI->isLegalInterleavedAccessType(SubVecTy, DL))
-      return Factor * TLI->getNumInterleavedAccesses(SubVecTy, DL);
+        TLI->isLegalInterleavedAccessType(Factor, SubVecTy, DL))
+      return Factor * BaseCost * TLI->getNumInterleavedAccesses(SubVecTy, DL);
+
+    // Some smaller than legal interleaved patterns are cheap as we can make
+    // use of the vmovn patterns to interleave a standard load. This is true
+    // for v4i8, v8i8 and v4i16 at least (but not for v4f16 as it is promoted
+    // differently).
+    if (ST->hasMVEIntegerOps() && Factor == 2 && NumElts / Factor > 2 &&
+        VecTy->isIntOrIntVectorTy() && DL.getTypeSizeInBits(SubVecTy) <= 128)
+      return 2 * BaseCost;
   }
 
   return BaseT::getInterleavedMemoryOpCost(Opcode, VecTy, Factor, Indices,
                                            Alignment, AddressSpace,
                                            UseMaskForCond, UseMaskForGaps);
 }
 
 bool ARMTTIImpl::isLoweredToCall(const Function *F) {

llvm/test/CodeGen/Thumb2/mve-vld2.ll

entry:
   %a = add <2 x i32> %s1, %s2
   store <2 x i32> %a, <2 x i32> *%dst
   ret void
 }
 
 define void @vld2_v4i32(<8 x i32> *%src, <4 x i32> *%dst) {
 ; CHECK-LABEL: vld2_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s8, s5
-; CHECK-NEXT:    vmov.f32 s9, s7
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov.f32 s6, s0
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s7, s2
-; CHECK-NEXT:    vadd.i32 q0, q1, q2
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <8 x i32>, <8 x i32>* %src, align 4
   %s1 = shufflevector <8 x i32> %l1, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %s2 = shufflevector <8 x i32> %l1, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %a = add <4 x i32> %s1, %s2
   store <4 x i32> %a, <4 x i32> *%dst
   ret void
 }
 
 define void @vld2_v8i32(<16 x i32> *%src, <8 x i32> *%dst) {
 ; CHECK-LABEL: vld2_v8i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.f32 s12, s1
-; CHECK-NEXT:    vmov.f32 s13, s3
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s2, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s3, s6
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vadd.i32 q0, q0, q3
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vmov.f32 s13, s11
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s10, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s11, s6
+; CHECK-NEXT:    add.w r2, r0, #32
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld20.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
 ; CHECK-NEXT:    vadd.i32 q1, q2, q3
-; CHECK-NEXT:    vstrw.32 q1, [r1]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x i32>, <16 x i32>* %src, align 4
   %s1 = shufflevector <16 x i32> %l1, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %s2 = shufflevector <16 x i32> %l1, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %a = add <8 x i32> %s1, %s2
   store <8 x i32> %a, <8 x i32> *%dst
   ret void
 }
 
 define void @vld2_v16i32(<32 x i32> *%src, <16 x i32> *%dst) {
 ; CHECK-LABEL: vld2_v16i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #96]
-; CHECK-NEXT:    vmov.f32 s8, s5
-; CHECK-NEXT:    vmov.f32 s9, s7
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov.f32 s6, s0
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s7, s2
-; CHECK-NEXT:    vadd.i32 q0, q1, q2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vmov.f32 s13, s11
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s10, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s11, s6
-; CHECK-NEXT:    vadd.i32 q1, q2, q3
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s20, s9
-; CHECK-NEXT:    vmov.f32 s21, s11
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s10, s12
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov.f32 s11, s14
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vadd.i32 q2, q2, q5
-; CHECK-NEXT:    vmov.f32 s20, s17
-; CHECK-NEXT:    vmov.f32 s21, s19
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s17, s18
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s18, s12
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov.f32 s19, s14
-; CHECK-NEXT:    vadd.i32 q3, q4, q5
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    add.w r12, r0, #96
+; CHECK-NEXT:    add.w r3, r0, #32
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
+; CHECK-NEXT:    vld20.32 {q1, q2}, [r2]
+; CHECK-NEXT:    vld20.32 {q3, q4}, [r12]
+; CHECK-NEXT:    vld20.32 {q5, q6}, [r3]
+; CHECK-NEXT:    vld21.32 {q5, q6}, [r3]
+; CHECK-NEXT:    vld21.32 {q1, q2}, [r2]
+; CHECK-NEXT:    vld21.32 {q3, q4}, [r12]
+; CHECK-NEXT:    @ kill: def $q1 killed $q1 killed $q1_q2
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vadd.i32 q5, q5, q6
+; CHECK-NEXT:    vadd.i32 q1, q1, q2
+; CHECK-NEXT:    vadd.i32 q3, q3, q4
+; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
 ; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vstrw.32 q5, [r1, #16]
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x i32>, <32 x i32>* %src, align 4
   %s1 = shufflevector <32 x i32> %l1, <32 x i32> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %s2 = shufflevector <32 x i32> %l1, <32 x i32> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %a = add <16 x i32> %s1, %s2
   store <16 x i32> %a, <16 x i32> *%dst
   ret void
entry:
   %a = add <4 x i16> %s1, %s2
   store <4 x i16> %a, <4 x i16> *%dst
   ret void
 }
 
 define void @vld2_v8i16(<16 x i16> *%src, <8 x i16> *%dst) {
 ; CHECK-LABEL: vld2_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.u16 r2, q1[1]
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[5]
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q3[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vadd.i16 q0, q3, q0
+; CHECK-NEXT:    vld20.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x i16>, <16 x i16>* %src, align 4
   %s1 = shufflevector <16 x i16> %l1, <16 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %s2 = shufflevector <16 x i16> %l1, <16 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %a = add <8 x i16> %s1, %s2
   store <8 x i16> %a, <8 x i16> *%dst
   ret void
 }
 
 define void @vld2_v16i16(<32 x i16> *%src, <16 x i16> *%dst) {
 ; CHECK-LABEL: vld2_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vmov.u16 r2, q1[1]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[5]
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[5]
-; CHECK-NEXT:    vmov.16 q0[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q0[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[0]
-; CHECK-NEXT:    vmov.16 q3[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[2]
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[4]
-; CHECK-NEXT:    vmov.16 q3[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[6]
-; CHECK-NEXT:    vmov.16 q3[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q3[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q3[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[4]
-; CHECK-NEXT:    vmov.16 q3[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q3[7], r2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vadd.i16 q0, q3, q0
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vmov.16 q1[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.16 q1[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[4]
-; CHECK-NEXT:    vmov.16 q1[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[6]
-; CHECK-NEXT:    vmov.16 q1[3], r2
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vadd.i16 q1, q1, q4
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    add.w r2, r0, #32
+; CHECK-NEXT:    vld20.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vld20.16 {q2, q3}, [r2]
+; CHECK-NEXT:    vld21.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.16 {q2, q3}, [r2]
+; CHECK-NEXT:    vadd.i16 q0, q0, q1
+; CHECK-NEXT:    vadd.i16 q1, q2, q3
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x i16>, <32 x i16>* %src, align 4
   %s1 = shufflevector <32 x i16> %l1, <32 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %s2 = shufflevector <32 x i16> %l1, <32 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %a = add <16 x i16> %s1, %s2
   store <16 x i16> %a, <16 x i16> *%dst
   ret void
entry:
   %a = add <8 x i8> %s1, %s2
   store <8 x i8> %a, <8 x i8> *%dst
   ret void
 }
 
 define void @vld2_v16i8(<32 x i8> *%src, <16 x i8> *%dst) {
 ; CHECK-LABEL: vld2_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.u8 r2, q1[1]
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q0[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[3]
-; CHECK-NEXT:    vmov.8 q0[1], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[5]
-; CHECK-NEXT:    vmov.8 q0[2], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    vmov.8 q0[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[9]
-; CHECK-NEXT:    vmov.8 q0[4], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[11]
-; CHECK-NEXT:    vmov.8 q0[5], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[13]
-; CHECK-NEXT:    vmov.8 q0[6], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[15]
-; CHECK-NEXT:    vmov.8 q0[7], r2
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q3[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q3[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q3[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q3[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q3[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q3[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q3[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q3[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q3[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q3[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q3[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q3[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q3[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q3[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q3[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q3[15], r0
-; CHECK-NEXT:    vadd.i8 q0, q3, q0
+; CHECK-NEXT:    vld20.8 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.8 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.i8 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x i8>, <32 x i8>* %src, align 4
   %s1 = shufflevector <32 x i8> %l1, <32 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %s2 = shufflevector <32 x i8> %l1, <32 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %a = add <16 x i8> %s1, %s2
   store <16 x i8> %a, <16 x i8> *%dst
entry:
   %a = fadd <2 x float> %s1, %s2
   store <2 x float> %a, <2 x float> *%dst
   ret void
 }
 
 define void @vld2_v4f32(<8 x float> *%src, <4 x float> *%dst) {
 ; CHECK-LABEL: vld2_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s8, s5
-; CHECK-NEXT:    vmov.f32 s9, s7
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov.f32 s6, s0
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s7, s2
-; CHECK-NEXT:    vadd.f32 q0, q1, q2
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <8 x float>, <8 x float>* %src, align 4
   %s1 = shufflevector <8 x float> %l1, <8 x float> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %s2 = shufflevector <8 x float> %l1, <8 x float> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %a = fadd <4 x float> %s1, %s2
   store <4 x float> %a, <4 x float> *%dst
   ret void
 }
 
 define void @vld2_v8f32(<16 x float> *%src, <8 x float> *%dst) {
 ; CHECK-LABEL: vld2_v8f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.f32 s12, s1
-; CHECK-NEXT:    vmov.f32 s13, s3
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s2, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s3, s6
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vadd.f32 q0, q0, q3
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vmov.f32 s13, s11
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s10, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s11, s6
+; CHECK-NEXT:    add.w r2, r0, #32
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld20.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vadd.f32 q1, q2, q3
-; CHECK-NEXT:    vstrw.32 q1, [r1]
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x float>, <16 x float>* %src, align 4
   %s1 = shufflevector <16 x float> %l1, <16 x float> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %s2 = shufflevector <16 x float> %l1, <16 x float> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %a = fadd <8 x float> %s1, %s2
   store <8 x float> %a, <8 x float> *%dst
   ret void
 }
 
 define void @vld2_v16f32(<32 x float> *%src, <16 x float> *%dst) {
 ; CHECK-LABEL: vld2_v16f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #96]
-; CHECK-NEXT:    vmov.f32 s8, s5
-; CHECK-NEXT:    vmov.f32 s9, s7
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov.f32 s6, s0
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s7, s2
-; CHECK-NEXT:    vadd.f32 q0, q1, q2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vmov.f32 s13, s11
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s10, s4
-; CHECK-NEXT:    vmov.f32 s15, s7
-; CHECK-NEXT:    vmov.f32 s11, s6
-; CHECK-NEXT:    vadd.f32 q1, q2, q3
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s20, s9
-; CHECK-NEXT:    vmov.f32 s21, s11
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s10, s12
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov.f32 s11, s14
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vadd.f32 q2, q2, q5
-; CHECK-NEXT:    vmov.f32 s20, s17
-; CHECK-NEXT:    vmov.f32 s21, s19
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s17, s18
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s18, s12
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov.f32 s19, s14
-; CHECK-NEXT:    vadd.f32 q3, q4, q5
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vld20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    add.w r12, r0, #96
+; CHECK-NEXT:    add.w r3, r0, #32
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld21.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
+; CHECK-NEXT:    vld20.32 {q1, q2}, [r2]
+; CHECK-NEXT:    vld20.32 {q3, q4}, [r12]
+; CHECK-NEXT:    vld20.32 {q5, q6}, [r3]
+; CHECK-NEXT:    vld21.32 {q5, q6}, [r3]
+; CHECK-NEXT:    vld21.32 {q1, q2}, [r2]
+; CHECK-NEXT:    vld21.32 {q3, q4}, [r12]
+; CHECK-NEXT:    @ kill: def $q1 killed $q1 killed $q1_q2
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vadd.f32 q5, q5, q6
+; CHECK-NEXT:    vadd.f32 q1, q1, q2
+; CHECK-NEXT:    vadd.f32 q3, q3, q4
+; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
 ; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vstrw.32 q5, [r1, #16]
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x float>, <32 x float>* %src, align 4
   %s1 = shufflevector <32 x float> %l1, <32 x float> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %s2 = shufflevector <32 x float> %l1, <32 x float> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %a = fadd <16 x float> %s1, %s2
   store <16 x float> %a, <16 x float> *%dst
   ret void
entry:
   %a = fadd <4 x half> %s1, %s2
   store <4 x half> %a, <4 x half> *%dst
   ret void
 }
 
 define void @vld2_v8f16(<16 x half> *%src, <8 x half> *%dst) {
 ; CHECK-LABEL: vld2_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8}
-; CHECK-NEXT:    vpush {d8}
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s12, s8
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov r3, s9
-; CHECK-NEXT:    vmov.16 q0[1], r3
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s9
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s8, s11
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s8, s4
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s8, s5
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s8, s6
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s8, s7
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vadd.f16 q0, q0, q3
+; CHECK-NEXT:    vld20.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vld21.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vadd.f16 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x half>, <16 x half>* %src, align 4
   %s1 = shufflevector <16 x half> %l1, <16 x half> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %s2 = shufflevector <16 x half> %l1, <16 x half> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %a = fadd <8 x half> %s1, %s2
   store <8 x half> %a, <8 x half> *%dst
   ret void
 }
 
 define void @vld2_v16f16(<32 x half> *%src, <16 x half> *%dst) {
 ; CHECK-LABEL: vld2_v16f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8}
-; CHECK-NEXT:    vpush {d8}
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s12, s8
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov r3, s9
-; CHECK-NEXT:    vmov.16 q0[1], r3
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.16 q0[6], r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmovx.f16 s12, s9
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov.16 q3[0], r2
-; CHECK-NEXT:    vmov.16 q3[1], r3
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmovx.f16 s8, s11
-; CHECK-NEXT:    vmov.16 q3[2], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s8, s4
-; CHECK-NEXT:    vmov.16 q3[3], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s8, s5
-; CHECK-NEXT:    vmov.16 q3[4], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s8, s6
-; CHECK-NEXT:    vmov.16 q3[5], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmovx.f16 s8, s7
-; CHECK-NEXT:    vmov.16 q3[6], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.16 q3[7], r2
-; CHECK-NEXT:    vmov r2, s7
-; CHECK-NEXT:    vmov.16 q0[7], r2
-; CHECK-NEXT:    vadd.f16 q1, q0, q3
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmovx.f16 s4, s8
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmovx.f16 s4, s9
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmovx.f16 s12, s10
-; CHECK-NEXT:    vmov.16 q1[1], r2
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s11
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s0
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s1
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s2
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s3
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r0, s11
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vadd.f16 q0, q3, q1
+; CHECK-NEXT:    add.w r2, r0, #32
+; CHECK-NEXT:    vld20.16 {q0, q1}, [r2]
+; CHECK-NEXT:    vld21.16 {q0, q1}, [r2]
+; CHECK-NEXT:    vadd.f16 q0, q0, q1
+; CHECK-NEXT:    vld20.16 {q1, q2}, [r0]
+; CHECK-NEXT:    vld21.16 {q1, q2}, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    vadd.f16 q0, q1, q2
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x half>, <32 x half>* %src, align 4
   %s1 = shufflevector <32 x half> %l1, <32 x half> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %s2 = shufflevector <32 x half> %l1, <32 x half> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %a = fadd <16 x half> %s1, %s2
   store <16 x half> %a, <16 x half> *%dst
   ret void

llvm/test/CodeGen/Thumb2/mve-vld4.ll

entry:
   %a3 = add <2 x i32> %a1, %a2
   store <2 x i32> %a3, <2 x i32> *%dst
   ret void
 }
 
 define void @vld4_v4i32(<16 x i32> *%src, <4 x i32> *%dst) {
 ; CHECK-LABEL: vld4_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vadd.i32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vadd.i32 q0, q0, q5
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i32 q4, q2, q3
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
 ; CHECK-NEXT:    vadd.i32 q0, q0, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x i32>, <16 x i32>* %src, align 4
   %s1 = shufflevector <16 x i32> %l1, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %s2 = shufflevector <16 x i32> %l1, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %s3 = shufflevector <16 x i32> %l1, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %s4 = shufflevector <16 x i32> %l1, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
   %a1 = add <4 x i32> %s1, %s2
   %a2 = add <4 x i32> %s3, %s4
   %a3 = add <4 x i32> %a1, %a2
   store <4 x i32> %a3, <4 x i32> *%dst
   ret void
 }
 
 define void @vld4_v8i32(<32 x i32> *%src, <8 x i32> *%dst) {
 ; CHECK-LABEL: vld4_v8i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .pad #88
+; CHECK-NEXT:    sub sp, #88
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vstrw.32 q5, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #64] @ 16-byte Reload
+; CHECK-NEXT:    vadd.i32 q4, q6, q7
+; CHECK-NEXT:    vadd.i32 q5, q1, q0
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
 ; CHECK-NEXT:    vadd.i32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstrw.32 q4, [r1]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i32 q5, q2, q3
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
 ; CHECK-NEXT:    vadd.i32 q0, q0, q5
-; CHECK-NEXT:    vmov.f64 d12, d3
-; CHECK-NEXT:    vadd.i32 q0, q0, q4
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
 ; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s23, s11
-; CHECK-NEXT:    vmov.f32 s25, s14
-; CHECK-NEXT:    vmov.f32 s20, s7
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vmov.f32 s26, s18
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s27, s10
-; CHECK-NEXT:    vadd.i32 q5, q6, q5
-; CHECK-NEXT:    vmov.f32 s26, s17
-; CHECK-NEXT:    vmov.f32 s27, s9
-; CHECK-NEXT:    vmov.f32 s24, s5
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s8
-; CHECK-NEXT:    vmov.f32 s25, s13
-; CHECK-NEXT:    vmov.f32 s5, s12
-; CHECK-NEXT:    vadd.i32 q1, q1, q6
-; CHECK-NEXT:    vadd.i32 q1, q1, q5
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    add sp, #88
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x i32>, <32 x i32>* %src, align 4
   %s1 = shufflevector <32 x i32> %l1, <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %s2 = shufflevector <32 x i32> %l1, <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %s3 = shufflevector <32 x i32> %l1, <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %s4 = shufflevector <32 x i32> %l1, <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
   %a1 = add <8 x i32> %s1, %s2
   %a2 = add <8 x i32> %s3, %s4
   %a3 = add <8 x i32> %a1, %a2
   store <8 x i32> %a3, <8 x i32> *%dst
   ret void
 }
 
 define void @vld4_v16i32(<64 x i32> *%src, <16 x i32> *%dst) {
 ; CHECK-LABEL: vld4_v16i32:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vadd.i32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #64]
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #112]
-; CHECK-NEXT:    vadd.i32 q0, q0, q5
-; CHECK-NEXT:    vadd.i32 q0, q0, q4
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #96]
-; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d0, d3
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vmov.f32 s26, s17
-; CHECK-NEXT:    vmov.f32 s23, s11
-; CHECK-NEXT:    vmov.f32 s27, s9
-; CHECK-NEXT:    vmov.f32 s20, s7
-; CHECK-NEXT:    vmov.f32 s24, s5
-; CHECK-NEXT:    vmov.f32 s1, s14
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #144]
-; CHECK-NEXT:    vmov.f32 s7, s8
-; CHECK-NEXT:    vmov.f32 s3, s10
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #128]
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s25, s13
-; CHECK-NEXT:    vadd.i32 q5, q0, q5
-; CHECK-NEXT:    vmov.f32 s5, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #176]
-; CHECK-NEXT:    vadd.i32 q0, q1, q6
-; CHECK-NEXT:    vadd.i32 q1, q0, q5
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #160]
-; CHECK-NEXT:    vmov.f64 d0, d5
-; CHECK-NEXT:    vmov.f32 s26, s23
-; CHECK-NEXT:    vmov.f32 s23, s14
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s27, s15
-; CHECK-NEXT:    vmov.f32 s31, s13
-; CHECK-NEXT:    vmov.f32 s24, s11
-; CHECK-NEXT:    vmov.f32 s28, s9
-; CHECK-NEXT:    vmov.f32 s1, s18
-; CHECK-NEXT:    vmov.f32 s10, s20
-; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vmov.f32 s11, s12
-; CHECK-NEXT:    vmov.f32 s3, s14
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #192]
-; CHECK-NEXT:    vmov.f32 s25, s19
-; CHECK-NEXT:    vmov.f32 s29, s17
-; CHECK-NEXT:    vadd.i32 q6, q0, q6
-; CHECK-NEXT:    vmov.f32 s9, s16
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #240]
-; CHECK-NEXT:    vadd.i32 q0, q2, q7
-; CHECK-NEXT:    vmov.f64 d10, d7
-; CHECK-NEXT:    vadd.i32 q2, q0, q6
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #224]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #208]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s30, s27
-; CHECK-NEXT:    vmov.f32 s31, s19
-; CHECK-NEXT:    vmov.f32 s21, s2
-; CHECK-NEXT:    vmov.f32 s28, s15
-; CHECK-NEXT:    vmov.f32 s27, s18
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s29, s3
-; CHECK-NEXT:    vmov.f32 s23, s18
-; CHECK-NEXT:    vadd.i32 q7, q5, q7
-; CHECK-NEXT:    vmov.f32 s22, s25
-; CHECK-NEXT:    vmov.f32 s23, s17
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s14, s24
-; CHECK-NEXT:    vmov.f32 s15, s16
-; CHECK-NEXT:    vmov.f32 s21, s1
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vadd.i32 q0, q3, q5
+; CHECK-NEXT:    .pad #152
+; CHECK-NEXT:    sub sp, #152
+; CHECK-NEXT:    add.w r2, r0, #128
+; CHECK-NEXT:    add r3, sp, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    add r4, sp, #64
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia r3, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add.w r3, r0, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    adds r0, #192
+; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vadd.i32 q4, q2, q3
+; CHECK-NEXT:    vmov q5, q0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q4, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vadd.i32 q4, q5, q0
+; CHECK-NEXT:    vldmia r4, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vstrw.32 q4, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    add r4, sp, #64
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia r4, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    add r3, sp, #64
+; CHECK-NEXT:    vstrw.32 q6, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vadd.i32 q4, q4, q5
+; CHECK-NEXT:    vadd.i32 q0, q1, q0
+; CHECK-NEXT:    vstrw.32 q0, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
 ; CHECK-NEXT:    vadd.i32 q0, q0, q7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vadd.i32 q0, q4, q0
+; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vldmia r3, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    add r0, sp, #64
+; CHECK-NEXT:    @ kill: def $q4 killed $q4 killed $q4_q5_q6_q7
+; CHECK-NEXT:    vstrw.32 q7, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov q2, q5
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vadd.i32 q4, q4, q2
+; CHECK-NEXT:    vadd.i32 q5, q6, q0
+; CHECK-NEXT:    vldmia r0, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i32 q4, q4, q5
+; CHECK-NEXT:    vadd.i32 q5, q2, q3
+; CHECK-NEXT:    vadd.i32 q0, q0, q1
+; CHECK-NEXT:    vstrw.32 q4, [r1, #48]
+; CHECK-NEXT:    vadd.i32 q0, q0, q5
+; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #16
+; CHECK-NEXT:    add sp, #152
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <64 x i32>, <64 x i32>* %src, align 4
   %s1 = shufflevector <64 x i32> %l1, <64 x i32> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x i32> %l1, <64 x i32> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x i32> %l1, <64 x i32> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x i32> %l1, <64 x i32> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = add <16 x i32> %s1, %s2
entry:
   %a3 = add <4 x i16> %a1, %a2
   store <4 x i16> %a3, <4 x i16> *%dst
   ret void
 }
 
 define void @vld4_v8i16(<32 x i16> *%src, <8 x i16> *%dst) {
 ; CHECK-LABEL: vld4_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
-; CHECK-NEXT:    vmov.u16 r2, q0[3]
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[7]
-; CHECK-NEXT:    vmov.16 q2[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q4[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q4[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vmov.16 q4[3], r2
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[4]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[5]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[6]
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[7]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vadd.i16 q4, q5, q4
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q7[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q7[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q7[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q7[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[1]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[2]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[3]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vadd.i16 q0, q5, q0
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vld40.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i16 q4, q2, q3
+; CHECK-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-NEXT:    vadd.i16 q0, q0, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x i16>, <32 x i16>* %src, align 4
   %s1 = shufflevector <32 x i16> %l1, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %s2 = shufflevector <32 x i16> %l1, <32 x i16> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %s3 = shufflevector <32 x i16> %l1, <32 x i16> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %s4 = shufflevector <32 x i16> %l1, <32 x i16> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
   %a1 = add <8 x i16> %s1, %s2
   %a2 = add <8 x i16> %s3, %s4
   %a3 = add <8 x i16> %a1, %a2
   store <8 x i16> %a3, <8 x i16> *%dst
   ret void
 }
 
 define void @vld4_v16i16(<64 x i16> *%src, <16 x i16> *%dst) {
 ; CHECK-LABEL: vld4_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vmov.u16 r2, q0[3]
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[7]
-; CHECK-NEXT:    vmov.16 q2[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q4[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q4[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #96]
-; CHECK-NEXT:    vmov.16 q4[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[3]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q5[4]
-; CHECK-NEXT:    vmov.16 q4[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q5[5]
-; CHECK-NEXT:    vmov.16 q4[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q5[6]
-; CHECK-NEXT:    vmov.16 q4[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q5[7]
-; CHECK-NEXT:    vmov.16 q4[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[2]
-; CHECK-NEXT:    vmov.16 q6[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[6]
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[2]
-; CHECK-NEXT:    vmov.16 q6[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[6]
-; CHECK-NEXT:    vmov.16 q6[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[0]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[3]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q6[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q6[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.16 q6[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[6]
-; CHECK-NEXT:    vmov.16 q6[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[6]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vmov.16 q6[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[0]
-; CHECK-NEXT:    vmov.16 q6[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[4]
-; CHECK-NEXT:    vmov.16 q6[3], r2
+; CHECK-NEXT:    .pad #88
+; CHECK-NEXT:    sub sp, #88
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld40.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld41.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.16 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld41.16 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld42.16 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld43.16 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vstrw.32 q5, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #64] @ 16-byte Reload
+; CHECK-NEXT:    vadd.i16 q4, q6, q7
+; CHECK-NEXT:    vadd.i16 q5, q1, q0
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
 ; CHECK-NEXT:    vadd.i16 q4, q5, q4
-; CHECK-NEXT:    vmov.u16 r2, q6[0]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[3]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q6[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[4]
-; CHECK-NEXT:    vmov.16 q6[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vmov.16 q6[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[4]
-; CHECK-NEXT:    vmov.16 q6[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[6]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q7[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[5]
-; CHECK-NEXT:    vmov.16 q7[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[1]
-; CHECK-NEXT:    vmov.16 q7[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[5]
-; CHECK-NEXT:    vmov.16 q7[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q7[0]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q7[1]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q7[2]
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q7[3]
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q1[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[5]
-; CHECK-NEXT:    vmov.16 q1[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[1]
-; CHECK-NEXT:    vmov.16 q1[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[5]
-; CHECK-NEXT:    vmov.16 q1[7], r2
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.u16 r2, q1[4]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[5]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[6]
-; CHECK-NEXT:    vmov.16 q0[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.16 q0[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[2]
-; CHECK-NEXT:    vmov.16 q3[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[6]
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q3[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q3[3], r2
-; CHECK-NEXT:    vadd.i16 q0, q5, q0
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vadd.i16 q0, q0, q4
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[3]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #48]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.u16 r0, q4[2]
-; CHECK-NEXT:    vmov.16 q6[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[6]
-; CHECK-NEXT:    vmov.16 q6[5], r2
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[6]
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q7[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q7[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q7[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q7[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[3]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q7[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q7[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q7[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q7[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vadd.i16 q5, q5, q7
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[1]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[5]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[0]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[4]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vadd.i16 q0, q1, q6
-; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vld42.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld43.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstrw.32 q4, [r1]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i16 q5, q2, q3
+; CHECK-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-NEXT:    vadd.i16 q0, q0, q5
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #16
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    add sp, #88
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <64 x i16>, <64 x i16>* %src, align 4
   %s1 = shufflevector <64 x i16> %l1, <64 x i16> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x i16> %l1, <64 x i16> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x i16> %l1, <64 x i16> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x i16> %l1, <64 x i16> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = add <16 x i16> %s1, %s2
entry:
   %a3 = add <8 x i8> %a1, %a2
   store <8 x i8> %a3, <8 x i8> *%dst
   ret void
 }
 
 define void @vld4_v16i8(<64 x i8> *%src, <16 x i8> *%dst) {
 ; CHECK-LABEL: vld4_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
-; CHECK-NEXT:    vmov.u8 r2, q0[3]
-; CHECK-NEXT:    vmov.8 q2[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q0[7]
-; CHECK-NEXT:    vmov.8 q2[1], r2
-; CHECK-NEXT:    vmov.u8 r2, q0[11]
-; CHECK-NEXT:    vmov.8 q2[2], r2
-; CHECK-NEXT:    vmov.u8 r2, q0[15]
-; CHECK-NEXT:    vmov.8 q2[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[3]
-; CHECK-NEXT:    vmov.8 q2[4], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    vmov.8 q2[5], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[11]
-; CHECK-NEXT:    vmov.8 q2[6], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[15]
-; CHECK-NEXT:    vmov.8 q2[7], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[0]
-; CHECK-NEXT:    vmov.8 q4[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[1]
-; CHECK-NEXT:    vmov.8 q4[1], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[2]
-; CHECK-NEXT:    vmov.8 q4[2], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[3]
-; CHECK-NEXT:    vmov.8 q4[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[4]
-; CHECK-NEXT:    vmov.8 q4[4], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[5]
-; CHECK-NEXT:    vmov.8 q4[5], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[6]
-; CHECK-NEXT:    vmov.8 q4[6], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[7]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vmov.8 q4[7], r2
-; CHECK-NEXT:    vmov.u8 r0, q3[3]
-; CHECK-NEXT:    vmov.u8 r2, q2[3]
-; CHECK-NEXT:    vmov.8 q5[8], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[7]
-; CHECK-NEXT:    vmov.8 q5[9], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[11]
-; CHECK-NEXT:    vmov.8 q5[10], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[15]
-; CHECK-NEXT:    vmov.8 q5[11], r2
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[7]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[11]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[15]
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[8]
-; CHECK-NEXT:    vmov.8 q4[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[9]
-; CHECK-NEXT:    vmov.8 q4[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[10]
-; CHECK-NEXT:    vmov.8 q4[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[11]
-; CHECK-NEXT:    vmov.8 q4[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[12]
-; CHECK-NEXT:    vmov.8 q4[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[13]
-; CHECK-NEXT:    vmov.8 q4[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[14]
-; CHECK-NEXT:    vmov.8 q4[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[15]
-; CHECK-NEXT:    vmov.8 q4[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[2]
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[6]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[10]
-; CHECK-NEXT:    vmov.8 q6[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[14]
-; CHECK-NEXT:    vmov.8 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q6[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q6[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[0]
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[1]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[2]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[3]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[4]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[5]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[6]
-; CHECK-NEXT:    vmov.8 q5[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[7]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q6[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q6[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q6[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[2]
-; CHECK-NEXT:    vmov.8 q6[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[6]
-; CHECK-NEXT:    vmov.8 q6[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[10]
-; CHECK-NEXT:    vmov.8 q6[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[14]
-; CHECK-NEXT:    vmov.8 q6[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[8]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[9]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[10]
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[11]
-; CHECK-NEXT:    vmov.8 q5[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[12]
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[13]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[14]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[15]
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[0]
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[4]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[8]
-; CHECK-NEXT:    vmov.8 q6[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[12]
-; CHECK-NEXT:    vmov.8 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q6[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q6[7], r0
-; CHECK-NEXT:    vadd.i8 q4, q5, q4
-; CHECK-NEXT:    vmov.u8 r0, q6[0]
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[1]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[2]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[3]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[4]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[5]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[6]
-; CHECK-NEXT:    vmov.8 q5[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[7]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q6[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q6[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q6[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[0]
-; CHECK-NEXT:    vmov.8 q6[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[4]
-; CHECK-NEXT:    vmov.8 q6[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[8]
-; CHECK-NEXT:    vmov.8 q6[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[12]
-; CHECK-NEXT:    vmov.8 q6[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[8]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[9]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[10]
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[11]
-; CHECK-NEXT:    vmov.8 q5[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[12]
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[13]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[14]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[1]
-; CHECK-NEXT:    vmov.8 q7[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[5]
-; CHECK-NEXT:    vmov.8 q7[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[9]
-; CHECK-NEXT:    vmov.8 q7[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[13]
-; CHECK-NEXT:    vmov.8 q7[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q7[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[5]
-; CHECK-NEXT:    vmov.8 q7[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q7[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q7[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[0]
-; CHECK-NEXT:    vmov.8 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[1]
-; CHECK-NEXT:    vmov.8 q0[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[2]
-; CHECK-NEXT:    vmov.8 q0[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[3]
-; CHECK-NEXT:    vmov.8 q0[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[4]
-; CHECK-NEXT:    vmov.8 q0[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[5]
-; CHECK-NEXT:    vmov.8 q0[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[6]
-; CHECK-NEXT:    vmov.8 q0[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[7]
-; CHECK-NEXT:    vmov.8 q0[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q1[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q1[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q1[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q1[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[1]
-; CHECK-NEXT:    vmov.8 q1[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[5]
-; CHECK-NEXT:    vmov.8 q1[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[9]
-; CHECK-NEXT:    vmov.8 q1[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[13]
-; CHECK-NEXT:    vmov.8 q1[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[15]
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vadd.i8 q0, q5, q0
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vld40.8 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.8 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.8 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.8 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.i8 q4, q2, q3
+; CHECK-NEXT:    vadd.i8 q0, q0, q1
 ; CHECK-NEXT:    vadd.i8 q0, q0, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <64 x i8>, <64 x i8>* %src, align 4
   %s1 = shufflevector <64 x i8> %l1, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x i8> %l1, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x i8> %l1, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x i8> %l1, <64 x i8> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = add <16 x i8> %s1, %s2
entry:
   %a3 = fadd <2 x float> %a1, %a2
   store <2 x float> %a3, <2 x float> *%dst
   ret void
 }
 
 define void @vld4_v4f32(<16 x float> *%src, <4 x float> *%dst) {
 ; CHECK-LABEL: vld4_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vadd.f32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vadd.f32 q0, q0, q5
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.f32 q4, q2, q3
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vadd.f32 q0, q0, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <16 x float>, <16 x float>* %src, align 4
   %s1 = shufflevector <16 x float> %l1, <16 x float> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %s2 = shufflevector <16 x float> %l1, <16 x float> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %s3 = shufflevector <16 x float> %l1, <16 x float> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %s4 = shufflevector <16 x float> %l1, <16 x float> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
   %a1 = fadd <4 x float> %s1, %s2
   %a2 = fadd <4 x float> %s3, %s4
   %a3 = fadd <4 x float> %a1, %a2
   store <4 x float> %a3, <4 x float> *%dst
   ret void
 }
 
 define void @vld4_v8f32(<32 x float> *%src, <8 x float> *%dst) {
 ; CHECK-LABEL: vld4_v8f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .pad #88
+; CHECK-NEXT:    sub sp, #88
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vstrw.32 q5, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #64] @ 16-byte Reload
+; CHECK-NEXT:    vadd.f32 q4, q6, q7
+; CHECK-NEXT:    vadd.f32 q5, q1, q0
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
 ; CHECK-NEXT:    vadd.f32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstrw.32 q4, [r1]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.f32 q5, q2, q3
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
 ; CHECK-NEXT:    vadd.f32 q0, q0, q5
-; CHECK-NEXT:    vmov.f64 d12, d3
-; CHECK-NEXT:    vadd.f32 q0, q0, q4
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
 ; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s23, s11
-; CHECK-NEXT:    vmov.f32 s25, s14
-; CHECK-NEXT:    vmov.f32 s20, s7
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vmov.f32 s26, s18
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s27, s10
-; CHECK-NEXT:    vadd.f32 q5, q6, q5
-; CHECK-NEXT:    vmov.f32 s26, s17
-; CHECK-NEXT:    vmov.f32 s27, s9
-; CHECK-NEXT:    vmov.f32 s24, s5
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s8
-; CHECK-NEXT:    vmov.f32 s25, s13
-; CHECK-NEXT:    vmov.f32 s5, s12
-; CHECK-NEXT:    vadd.f32 q1, q1, q6
-; CHECK-NEXT:    vadd.f32 q1, q1, q5
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    add sp, #88
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x float>, <32 x float>* %src, align 4
   %s1 = shufflevector <32 x float> %l1, <32 x float> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %s2 = shufflevector <32 x float> %l1, <32 x float> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %s3 = shufflevector <32 x float> %l1, <32 x float> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %s4 = shufflevector <32 x float> %l1, <32 x float> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
   %a1 = fadd <8 x float> %s1, %s2
   %a2 = fadd <8 x float> %s3, %s4
   %a3 = fadd <8 x float> %a1, %a2
   store <8 x float> %a3, <8 x float> *%dst
   ret void
 }
 
 define void @vld4_v16f32(<64 x float> *%src, <16 x float> *%dst) {
 ; CHECK-LABEL: vld4_v16f32:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f64 d10, d1
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vmov.f32 s21, s10
-; CHECK-NEXT:    vmov.f32 s16, s3
-; CHECK-NEXT:    vmov.f32 s15, s6
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s23, s6
-; CHECK-NEXT:    vadd.f32 q4, q5, q4
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s5
-; CHECK-NEXT:    vmov.f32 s20, s1
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s3, s4
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #64]
-; CHECK-NEXT:    vmov.f32 s21, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #112]
-; CHECK-NEXT:    vadd.f32 q0, q0, q5
-; CHECK-NEXT:    vadd.f32 q0, q0, q4
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #96]
-; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d0, d3
-; CHECK-NEXT:    vmov.f32 s22, s19
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vmov.f32 s26, s17
-; CHECK-NEXT:    vmov.f32 s23, s11
-; CHECK-NEXT:    vmov.f32 s27, s9
-; CHECK-NEXT:    vmov.f32 s20, s7
-; CHECK-NEXT:    vmov.f32 s24, s5
-; CHECK-NEXT:    vmov.f32 s1, s14
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #144]
-; CHECK-NEXT:    vmov.f32 s7, s8
-; CHECK-NEXT:    vmov.f32 s3, s10
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #128]
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s25, s13
-; CHECK-NEXT:    vadd.f32 q5, q0, q5
-; CHECK-NEXT:    vmov.f32 s5, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #176]
-; CHECK-NEXT:    vadd.f32 q0, q1, q6
-; CHECK-NEXT:    vadd.f32 q1, q0, q5
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #160]
-; CHECK-NEXT:    vmov.f64 d0, d5
-; CHECK-NEXT:    vmov.f32 s26, s23
-; CHECK-NEXT:    vmov.f32 s23, s14
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s27, s15
-; CHECK-NEXT:    vmov.f32 s31, s13
-; CHECK-NEXT:    vmov.f32 s24, s11
-; CHECK-NEXT:    vmov.f32 s28, s9
-; CHECK-NEXT:    vmov.f32 s1, s18
-; CHECK-NEXT:    vmov.f32 s10, s20
-; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vmov.f32 s11, s12
-; CHECK-NEXT:    vmov.f32 s3, s14
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #192]
-; CHECK-NEXT:    vmov.f32 s25, s19
-; CHECK-NEXT:    vmov.f32 s29, s17
-; CHECK-NEXT:    vadd.f32 q6, q0, q6
-; CHECK-NEXT:    vmov.f32 s9, s16
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #240]
-; CHECK-NEXT:    vadd.f32 q0, q2, q7
-; CHECK-NEXT:    vmov.f64 d10, d7
-; CHECK-NEXT:    vadd.f32 q2, q0, q6
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #224]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #208]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s30, s27
-; CHECK-NEXT:    vmov.f32 s31, s19
-; CHECK-NEXT:    vmov.f32 s21, s2
-; CHECK-NEXT:    vmov.f32 s28, s15
-; CHECK-NEXT:    vmov.f32 s27, s18
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s29, s3
-; CHECK-NEXT:    vmov.f32 s23, s18
-; CHECK-NEXT:    vadd.f32 q7, q5, q7
-; CHECK-NEXT:    vmov.f32 s22, s25
-; CHECK-NEXT:    vmov.f32 s23, s17
-; CHECK-NEXT:    vmov.f32 s20, s13
-; CHECK-NEXT:    vmov.f32 s14, s24
-; CHECK-NEXT:    vmov.f32 s15, s16
-; CHECK-NEXT:    vmov.f32 s21, s1
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vadd.f32 q0, q3, q5
+; CHECK-NEXT:    .pad #152
+; CHECK-NEXT:    sub sp, #152
+; CHECK-NEXT:    add.w r2, r0, #128
+; CHECK-NEXT:    add r3, sp, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    add r4, sp, #64
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia r3, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add.w r3, r0, #64
+; CHECK-NEXT:    vld40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    adds r0, #192
+; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vadd.f32 q4, q2, q3
+; CHECK-NEXT:    vmov q5, q0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q4, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vadd.f32 q4, q5, q0
+; CHECK-NEXT:    vldmia r4, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vstrw.32 q4, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    add r4, sp, #64
+; CHECK-NEXT:    vld42.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vstmia r4, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    add r3, sp, #64
+; CHECK-NEXT:    vstrw.32 q6, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vadd.f32 q4, q4, q5
+; CHECK-NEXT:    vadd.f32 q0, q1, q0
+; CHECK-NEXT:    vstrw.32 q0, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
 ; CHECK-NEXT:    vadd.f32 q0, q0, q7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vadd.f32 q0, q4, q0
+; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vldmia r3, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vld43.32 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vld40.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld41.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld42.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    vld43.32 {q4, q5, q6, q7}, [r0]
+; CHECK-NEXT:    add r0, sp, #64
+; CHECK-NEXT:    @ kill: def $q4 killed $q4 killed $q4_q5_q6_q7
+; CHECK-NEXT:    vstrw.32 q7, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov q2, q5
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vadd.f32 q4, q4, q2
+; CHECK-NEXT:    vadd.f32 q5, q6, q0
+; CHECK-NEXT:    vldmia r0, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.f32 q4, q4, q5
+; CHECK-NEXT:    vadd.f32 q5, q2, q3
+; CHECK-NEXT:    vadd.f32 q0, q0, q1
+; CHECK-NEXT:    vstrw.32 q4, [r1, #48]
+; CHECK-NEXT:    vadd.f32 q0, q0, q5
+; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #16
+; CHECK-NEXT:    add sp, #152
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <64 x float>, <64 x float>* %src, align 4
   %s1 = shufflevector <64 x float> %l1, <64 x float> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x float> %l1, <64 x float> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x float> %l1, <64 x float> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x float> %l1, <64 x float> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = fadd <16 x float> %s1, %s2
entry:
   %a3 = fadd <4 x half> %a1, %a2
   store <4 x half> %a3, <4 x half> *%dst
   ret void
 }
 
 define void @vld4_v8f16(<32 x half> *%src, <8 x half> *%dst) {
 ; CHECK-LABEL: vld4_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    vmovx.f16 s20, s5
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov r3, s7
-; CHECK-NEXT:    vmov.16 q4[1], r3
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    vmov.16 q4[2], r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov.16 q4[3], r2
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov.16 q4[4], r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vmov.16 q4[5], r2
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s7
-; CHECK-NEXT:    vmovx.f16 s24, s1
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmovx.f16 s24, s3
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s9
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s11
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s13
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s15
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s0
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov r0, s15
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vadd.f16 q4, q4, q5
-; CHECK-NEXT:    vmovx.f16 s20, s4
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s6
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s2
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s8
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s10
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s12
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s24, s14
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.16 q1[1], r2
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vadd.f16 q0, q1, q5
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vld40.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld41.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld42.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vld43.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.f16 q4, q2, q3
+; CHECK-NEXT:    vadd.f16 q0, q0, q1
 ; CHECK-NEXT:    vadd.f16 q0, q0, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <32 x half>, <32 x half>* %src, align 4
   %s1 = shufflevector <32 x half> %l1, <32 x half> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %s2 = shufflevector <32 x half> %l1, <32 x half> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %s3 = shufflevector <32 x half> %l1, <32 x half> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %s4 = shufflevector <32 x half> %l1, <32 x half> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
   %a1 = fadd <8 x half> %s1, %s2
   %a2 = fadd <8 x half> %s3, %s4
   %a3 = fadd <8 x half> %a1, %a2
   store <8 x half> %a3, <8 x half> *%dst
   ret void
 }
 
 define void @vld4_v16f16(<64 x half> *%src, <16 x half> *%dst) {
 ; CHECK-LABEL: vld4_v16f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12}
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    vmovx.f16 s20, s5
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov r3, s7
-; CHECK-NEXT:    vmov.16 q4[1], r3
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    vmov.16 q4[2], r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov.16 q4[3], r2
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov.16 q4[4], r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vmov.16 q4[5], r2
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    vmov.16 q4[6], r2
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmovx.f16 s20, s7
-; CHECK-NEXT:    vmovx.f16 s24, s1
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmov.16 q5[1], r3
-; CHECK-NEXT:    vmovx.f16 s24, s3
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s9
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s11
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s13
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s15
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s0
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    vmov.16 q4[7], r2
-; CHECK-NEXT:    vadd.f16 q4, q4, q5
-; CHECK-NEXT:    vmovx.f16 s20, s4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmovx.f16 s20, s6
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.16 q5[1], r3
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s2
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s8
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s10
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s12
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s14
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov r2, s24
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vmov.16 q1[0], r2
-; CHECK-NEXT:    vmov.16 q1[1], r3
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov.16 q1[2], r2
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.16 q1[3], r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov.16 q1[4], r2
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov.16 q1[5], r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov.16 q1[6], r2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.16 q1[7], r2
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vadd.f16 q0, q1, q5
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vadd.f16 q4, q0, q4
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vstrw.32 q4, [r1, #16]
-; CHECK-NEXT:    vmovx.f16 s16, s13
-; CHECK-NEXT:    vmovx.f16 s20, s9
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s15
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.16 q4[1], r2
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s11
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s5
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s7
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s1
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s3
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s24, s12
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov r0, s11
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vadd.f16 q4, q5, q4
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmovx.f16 s12, s14
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmovx.f16 s24, s8
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmovx.f16 s8, s10
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s8, s4
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s4, s6
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmovx.f16 s4, s0
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmovx.f16 s4, s2
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vadd.f16 q0, q5, q3
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    add.w r2, r0, #64
+; CHECK-NEXT:    vld40.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld41.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld42.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    vld43.16 {q0, q1, q2, q3}, [r2]
+; CHECK-NEXT:    @ kill: def $q0 killed $q0 killed $q0_q1_q2_q3
+; CHECK-NEXT:    vadd.f16 q4, q2, q3
+; CHECK-NEXT:    vadd.f16 q0, q0, q1
 ; CHECK-NEXT:    vadd.f16 q0, q0, q4
+; CHECK-NEXT:    vld40.16 {q1, q2, q3, q4}, [r0]
+; CHECK-NEXT:    vld41.16 {q1, q2, q3, q4}, [r0]
+; CHECK-NEXT:    vld42.16 {q1, q2, q3, q4}, [r0]
+; CHECK-NEXT:    vld43.16 {q1, q2, q3, q4}, [r0]
+; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-NEXT:    @ kill: def $q1 killed $q1 killed $q1_q2_q3_q4
+; CHECK-NEXT:    vadd.f16 q0, q3, q4
+; CHECK-NEXT:    vadd.f16 q1, q1, q2
+; CHECK-NEXT:    vadd.f16 q0, q1, q0
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <64 x half>, <64 x half>* %src, align 4
   %s1 = shufflevector <64 x half> %l1, <64 x half> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x half> %l1, <64 x half> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x half> %l1, <64 x half> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x half> %l1, <64 x half> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = fadd <16 x half> %s1, %s2

llvm/test/CodeGen/Thumb2/mve-vst2.ll

entry:
   %s = shufflevector <2 x i32> %l1, <2 x i32> %l2, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   store <4 x i32> %s, <4 x i32> *%dst
   ret void
 }
 
 define void @vst2_v4i32(<4 x i32> *%src, <8 x i32> *%dst) {
 ; CHECK-LABEL: vst2_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d4, d3
-; CHECK-NEXT:    vmov.f64 d6, d2
-; CHECK-NEXT:    vmov.f32 s9, s2
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vmov.f32 s10, s7
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s15, s1
-; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q3, [r1]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <4 x i32>, <4 x i32>* %src, i32 0
   %l1 = load <4 x i32>, <4 x i32>* %s1, align 4
   %s2 = getelementptr <4 x i32>, <4 x i32>* %src, i32 1
   %l2 = load <4 x i32>, <4 x i32>* %s2, align 4
   %s = shufflevector <4 x i32> %l1, <4 x i32> %l2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   store <8 x i32> %s, <8 x i32> *%dst
   ret void
 }
 
 define void @vst2_v8i32(<8 x i32> *%src, <16 x i32> *%dst) {
 ; CHECK-LABEL: vst2_v8i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q4, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmov.f64 d6, d8
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vmov.f32 s14, s17
-; CHECK-NEXT:    vmov.f32 s15, s1
-; CHECK-NEXT:    vmov.f32 s0, s18
-; CHECK-NEXT:    vstrw.32 q3, [r1]
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s2, s19
-; CHECK-NEXT:    vmov.f64 d8, d4
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmov.f32 s18, s9
-; CHECK-NEXT:    vmov.f32 s19, s5
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vstrw.32 q4, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s6, s11
-; CHECK-NEXT:    vstrw.32 q1, [r1, #48]
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    add.w r0, r1, #32
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst20.32 {q2, q3}, [r0]
+; CHECK-NEXT:    vst21.32 {q2, q3}, [r0]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i32>, <8 x i32>* %src, i32 0
   %l1 = load <8 x i32>, <8 x i32>* %s1, align 4
   %s2 = getelementptr <8 x i32>, <8 x i32>* %src, i32 1
   %l2 = load <8 x i32>, <8 x i32>* %s2, align 4
   %s = shufflevector <8 x i32> %l1, <8 x i32> %l2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x i32> %s, <16 x i32> *%dst
   ret void
 }
 
 define void @vst2_v16i32(<16 x i32> *%src, <32 x i32> *%dst) {
 ; CHECK-LABEL: vst2_v16i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #32
-; CHECK-NEXT:    sub sp, #32
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
-; CHECK-NEXT:    vstrw.32 q7, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d14, d0
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vmov.f32 s29, s16
-; CHECK-NEXT:    vmov.f32 s30, s1
-; CHECK-NEXT:    vmov.f32 s31, s17
-; CHECK-NEXT:    vstrw.32 q7, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d14, d10
-; CHECK-NEXT:    vmov.f32 s29, s4
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s31, s5
-; CHECK-NEXT:    vmov.f32 s4, s22
-; CHECK-NEXT:    vstrw.32 q7, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s6, s23
-; CHECK-NEXT:    vmov.f64 d10, d12
-; CHECK-NEXT:    vstrw.32 q1, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s16, s2
-; CHECK-NEXT:    vmov.f32 s21, s8
-; CHECK-NEXT:    vmov.f32 s17, s18
-; CHECK-NEXT:    vmov.f32 s22, s25
-; CHECK-NEXT:    vmov.f32 s18, s3
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s23, s9
-; CHECK-NEXT:    vstrw.32 q4, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s8, s26
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s10, s27
-; CHECK-NEXT:    vmov.f64 d12, d0
-; CHECK-NEXT:    vstrw.32 q2, [r1, #80]
-; CHECK-NEXT:    vmov.f32 s25, s12
-; CHECK-NEXT:    vmov.f32 s26, s1
-; CHECK-NEXT:    vmov.f32 s27, s13
-; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vstrw.32 q6, [r1, #96]
-; CHECK-NEXT:    vmov.f32 s13, s14
-; CHECK-NEXT:    vmov.f32 s14, s3
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q3, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #32
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #96]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q7, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q6, [r0]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
+; CHECK-NEXT:    add.w r0, r1, #96
+; CHECK-NEXT:    add.w r2, r1, #64
+; CHECK-NEXT:    add.w r3, r1, #32
+; CHECK-NEXT:    vst20.32 {q6, q7}, [r1]
+; CHECK-NEXT:    vst21.32 {q6, q7}, [r1]
+; CHECK-NEXT:    vst20.32 {q4, q5}, [r3]
+; CHECK-NEXT:    vst21.32 {q4, q5}, [r3]
+; CHECK-NEXT:    vst20.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vst21.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r0]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i32>, <16 x i32>* %src, i32 0
   %l1 = load <16 x i32>, <16 x i32>* %s1, align 4
   %s2 = getelementptr <16 x i32>, <16 x i32>* %src, i32 1
   %l2 = load <16 x i32>, <16 x i32>* %s2, align 4
   %s = shufflevector <16 x i32> %l1, <16 x i32> %l2, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
entry:
   %s = shufflevector <4 x i16> %l1, <4 x i16> %l2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   store <8 x i16> %s, <8 x i16> *%dst
   ret void
 }
 
 define void @vst2_v8i16(<8 x i16> *%src, <16 x i16> *%dst) {
 ; CHECK-LABEL: vst2_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.u16 r2, q1[4]
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q3[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q3, [r1]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vst20.16 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.16 {q0, q1}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i16>, <8 x i16>* %src, i32 0
   %l1 = load <8 x i16>, <8 x i16>* %s1, align 4
   %s2 = getelementptr <8 x i16>, <8 x i16>* %src, i32 1
   %l2 = load <8 x i16>, <8 x i16>* %s2, align 4
   %s = shufflevector <8 x i16> %l1, <8 x i16> %l2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x i16> %s, <16 x i16> *%dst
   ret void
 }
 
 define void @vst2_v16i16(<16 x i16> *%src, <32 x i16> *%dst) {
 ; CHECK-LABEL: vst2_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #48]
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q0[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[1]
-; CHECK-NEXT:    vmov.16 q0[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q0[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[3]
-; CHECK-NEXT:    vmov.16 q0[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[4]
-; CHECK-NEXT:    vmov.16 q1[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[4]
-; CHECK-NEXT:    vmov.16 q1[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[5]
-; CHECK-NEXT:    vmov.16 q1[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[5]
-; CHECK-NEXT:    vmov.16 q1[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q1[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[6]
-; CHECK-NEXT:    vmov.16 q1[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q1[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[7]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.16 q1[7], r2
-; CHECK-NEXT:    vmov.u16 r0, q4[0]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[1]
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[2]
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[3]
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[4]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[5]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[6]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vstrw.32 q5, [r1, #48]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    add.w r0, r1, #32
+; CHECK-NEXT:    vst20.16 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.16 {q0, q1}, [r1]
+; CHECK-NEXT:    vst20.16 {q2, q3}, [r0]
+; CHECK-NEXT:    vst21.16 {q2, q3}, [r0]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i16>, <16 x i16>* %src, i32 0
   %l1 = load <16 x i16>, <16 x i16>* %s1, align 4
   %s2 = getelementptr <16 x i16>, <16 x i16>* %src, i32 1
   %l2 = load <16 x i16>, <16 x i16>* %s2, align 4
   %s = shufflevector <16 x i16> %l1, <16 x i16> %l2, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   store <32 x i16> %s, <32 x i16> *%dst
entry:
   %s = shufflevector <8 x i8> %l1, <8 x i8> %l2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x i8> %s, <16 x i8> *%dst
   ret void
 }
 
 define void @vst2_v16i8(<16 x i8> *%src, <32 x i8> *%dst) {
 ; CHECK-LABEL: vst2_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.u8 r2, q1[8]
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q0[0], r2
-; CHECK-NEXT:    vmov.8 q0[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q0[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q0[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q0[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q0[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q0[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q0[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q3[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q3[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q3[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q3[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q3[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q3[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.8 q3[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q3[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q3[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q3[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[5]
-; CHECK-NEXT:    vmov.8 q3[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q3[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q3[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q3[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[7]
-; CHECK-NEXT:    vmov.8 q3[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q3[15], r0
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q3, [r1]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vst20.8 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.8 {q0, q1}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i8>, <16 x i8>* %src, i32 0
   %l1 = load <16 x i8>, <16 x i8>* %s1, align 4
   %s2 = getelementptr <16 x i8>, <16 x i8>* %src, i32 1
   %l2 = load <16 x i8>, <16 x i8>* %s2, align 4
   %s = shufflevector <16 x i8> %l1, <16 x i8> %l2, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   store <32 x i8> %s, <32 x i8> *%dst
entry:
   %s = shufflevector <2 x float> %l1, <2 x float> %l2, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   store <4 x float> %s, <4 x float> *%dst
   ret void
 }
 
 define void @vst2_v4f32(<4 x float> *%src, <8 x float> *%dst) {
 ; CHECK-LABEL: vst2_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d4, d3
-; CHECK-NEXT:    vmov.f64 d6, d2
-; CHECK-NEXT:    vmov.f32 s9, s2
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vmov.f32 s10, s7
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s15, s1
-; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
-; CHECK-NEXT:    vstrw.32 q3, [r1]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <4 x float>, <4 x float>* %src, i32 0
   %l1 = load <4 x float>, <4 x float>* %s1, align 4
   %s2 = getelementptr <4 x float>, <4 x float>* %src, i32 1
   %l2 = load <4 x float>, <4 x float>* %s2, align 4
   %s = shufflevector <4 x float> %l1, <4 x float> %l2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   store <8 x float> %s, <8 x float> *%dst
   ret void
 }
 
 define void @vst2_v8f32(<8 x float> *%src, <16 x float> *%dst) {
 ; CHECK-LABEL: vst2_v8f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q4, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmov.f64 d6, d8
-; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vmov.f32 s14, s17
-; CHECK-NEXT:    vmov.f32 s15, s1
-; CHECK-NEXT:    vmov.f32 s0, s18
-; CHECK-NEXT:    vstrw.32 q3, [r1]
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s2, s19
-; CHECK-NEXT:    vmov.f64 d8, d4
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmov.f32 s18, s9
-; CHECK-NEXT:    vmov.f32 s19, s5
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vstrw.32 q4, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s6, s11
-; CHECK-NEXT:    vstrw.32 q1, [r1, #48]
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    add.w r0, r1, #32
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r1]
+; CHECK-NEXT:    vst20.32 {q2, q3}, [r0]
+; CHECK-NEXT:    vst21.32 {q2, q3}, [r0]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x float>, <8 x float>* %src, i32 0
   %l1 = load <8 x float>, <8 x float>* %s1, align 4
   %s2 = getelementptr <8 x float>, <8 x float>* %src, i32 1
   %l2 = load <8 x float>, <8 x float>* %s2, align 4
   %s = shufflevector <8 x float> %l1, <8 x float> %l2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x float> %s, <16 x float> *%dst
   ret void
 }
 
 define void @vst2_v16f32(<16 x float> *%src, <32 x float> *%dst) {
 ; CHECK-LABEL: vst2_v16f32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #32
-; CHECK-NEXT:    sub sp, #32
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
-; CHECK-NEXT:    vstrw.32 q7, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d14, d0
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
-; CHECK-NEXT:    vmov.f32 s29, s16
-; CHECK-NEXT:    vmov.f32 s30, s1
-; CHECK-NEXT:    vmov.f32 s31, s17
-; CHECK-NEXT:    vstrw.32 q7, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d14, d10
-; CHECK-NEXT:    vmov.f32 s29, s4
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s31, s5
-; CHECK-NEXT:    vmov.f32 s4, s22
-; CHECK-NEXT:    vstrw.32 q7, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s6, s23
-; CHECK-NEXT:    vmov.f64 d10, d12
-; CHECK-NEXT:    vstrw.32 q1, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s16, s2
-; CHECK-NEXT:    vmov.f32 s21, s8
-; CHECK-NEXT:    vmov.f32 s17, s18
-; CHECK-NEXT:    vmov.f32 s22, s25
-; CHECK-NEXT:    vmov.f32 s18, s3
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s23, s9
-; CHECK-NEXT:    vstrw.32 q4, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s8, s26
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s10, s27
-; CHECK-NEXT:    vmov.f64 d12, d0
-; CHECK-NEXT:    vstrw.32 q2, [r1, #80]
-; CHECK-NEXT:    vmov.f32 s25, s12
-; CHECK-NEXT:    vmov.f32 s26, s1
-; CHECK-NEXT:    vmov.f32 s27, s13
-; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vstrw.32 q6, [r1, #96]
-; CHECK-NEXT:    vmov.f32 s13, s14
-; CHECK-NEXT:    vmov.f32 s14, s3
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q3, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #32
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #96]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q7, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q6, [r0]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
+; CHECK-NEXT:    add.w r0, r1, #96
+; CHECK-NEXT:    add.w r2, r1, #64
+; CHECK-NEXT:    add.w r3, r1, #32
+; CHECK-NEXT:    vst20.32 {q6, q7}, [r1]
+; CHECK-NEXT:    vst21.32 {q6, q7}, [r1]
+; CHECK-NEXT:    vst20.32 {q4, q5}, [r3]
+; CHECK-NEXT:    vst21.32 {q4, q5}, [r3]
+; CHECK-NEXT:    vst20.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vst21.32 {q2, q3}, [r2]
+; CHECK-NEXT:    vst20.32 {q0, q1}, [r0]
+; CHECK-NEXT:    vst21.32 {q0, q1}, [r0]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x float>, <16 x float>* %src, i32 0
   %l1 = load <16 x float>, <16 x float>* %s1, align 4
   %s2 = getelementptr <16 x float>, <16 x float>* %src, i32 1
   %l2 = load <16 x float>, <16 x float>* %s2, align 4
   %s = shufflevector <16 x float> %l1, <16 x float> %l2, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
entry:
   %s = shufflevector <4 x half> %l1, <4 x half> %l2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   store <8 x half> %s, <8 x half> *%dst
   ret void
 }
 
 define void @vst2_v8f16(<8 x half> *%src, <16 x half> *%dst) {
 ; CHECK-LABEL: vst2_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmovx.f16 s12, s6
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s2
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s7
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s3
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s4
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmovx.f16 s4, s5
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s0
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s0, s1
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vstrw.32 q2, [r1]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vst20.16 {q0, q1}, [r1]
+; CHECK-NEXT:    vst21.16 {q0, q1}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x half>, <8 x half>* %src, i32 0
   %l1 = load <8 x half>, <8 x half>* %s1, align 4
   %s2 = getelementptr <8 x half>, <8 x half>* %src, i32 1
   %l2 = load <8 x half>, <8 x half>* %s2, align 4
   %s = shufflevector <8 x half> %l1, <8 x half> %l2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x half> %s, <16 x half> *%dst
   ret void
 }
 
 define void @vst2_v16f16(<16 x half> *%src, <32 x half> *%dst) {
 ; CHECK-LABEL: vst2_v16f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10}
-; CHECK-NEXT:    vpush {d8, d9, d10}
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmovx.f16 s0, s12
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov.16 q4[1], r3
-; CHECK-NEXT:    vmovx.f16 s0, s8
-; CHECK-NEXT:    vmov.16 q4[2], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmovx.f16 s0, s13
-; CHECK-NEXT:    vmov.16 q4[3], r2
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    vmov.16 q4[4], r2
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov.16 q4[5], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmovx.f16 s0, s9
-; CHECK-NEXT:    vmov.16 q4[6], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
-; CHECK-NEXT:    vmov.16 q4[7], r2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vstrw.32 q4, [r1, #32]
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmovx.f16 s20, s14
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s20, s10
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmovx.f16 s12, s15
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov r0, s15
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov r0, s11
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s8, s11
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmovx.f16 s12, s4
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q2[0], r0
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov.16 q2[1], r2
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s0
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s5
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s1
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s6
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vstrw.32 q2, [r1]
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q2[0], r2
-; CHECK-NEXT:    vmovx.f16 s4, s7
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s2
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s0, s3
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vstrw.32 q4, [r1, #48]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
-; CHECK-NEXT:    vpop {d8, d9, d10}
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q2, [r0]
+; CHECK-NEXT:    add.w r0, r1, #32
+; CHECK-NEXT:    vst20.16 {q2, q3}, [r1]
+; CHECK-NEXT:    vst21.16 {q2, q3}, [r1]
+; CHECK-NEXT:    vst20.16 {q0, q1}, [r0]
+; CHECK-NEXT:    vst21.16 {q0, q1}, [r0]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x half>, <16 x half>* %src, i32 0
   %l1 = load <16 x half>, <16 x half>* %s1, align 4
   %s2 = getelementptr <16 x half>, <16 x half>* %src, i32 1
   %l2 = load <16 x half>, <16 x half>* %s2, align 4
   %s = shufflevector <16 x half> %l1, <16 x half> %l2, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   store <32 x half> %s, <32 x half> *%dst

llvm/test/CodeGen/Thumb2/mve-vst4.ll

entry:
   %s = shufflevector <4 x i32> %t1, <4 x i32> %t2, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
   store <8 x i32> %s, <8 x i32> *%dst
   ret void
 }
 
 define void @vst4_v4i32(<4 x i32> *%src, <16 x i32> *%dst) {
 ; CHECK-LABEL: vst4_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s0, s9
-; CHECK-NEXT:    vmov.32 r0, q3[1]
-; CHECK-NEXT:    vdup.32 q4, r0
-; CHECK-NEXT:    vmov.f32 s1, s5
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vmov.32 r0, q3[0]
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vdup.32 q4, r0
-; CHECK-NEXT:    vmov.f32 s9, s4
-; CHECK-NEXT:    vmov.32 r0, q3[3]
-; CHECK-NEXT:    vmov.f32 s16, s8
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s20, s11
-; CHECK-NEXT:    vmov.32 r0, q3[2]
-; CHECK-NEXT:    vmov.f32 s8, s10
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s21, s7
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmov.f32 s9, s6
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vstrw.32 q4, [r1]
-; CHECK-NEXT:    vmov.f32 s10, s6
-; CHECK-NEXT:    vmov.f32 s23, s27
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vstrw.32 q5, [r1, #48]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <4 x i32>, <4 x i32>* %src, i32 0
   %l1 = load <4 x i32>, <4 x i32>* %s1, align 4
   %s2 = getelementptr <4 x i32>, <4 x i32>* %src, i32 1
   %l2 = load <4 x i32>, <4 x i32>* %s2, align 4
   %s3 = getelementptr <4 x i32>, <4 x i32>* %src, i32 2
   %l3 = load <4 x i32>, <4 x i32>* %s3, align 4
   %s4 = getelementptr <4 x i32>, <4 x i32>* %src, i32 3
   %l4 = load <4 x i32>, <4 x i32>* %s3, align 4
   %t1 = shufflevector <4 x i32> %l1, <4 x i32> %l2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %t2 = shufflevector <4 x i32> %l3, <4 x i32> %l4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %s = shufflevector <8 x i32> %t1, <8 x i32> %t2, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
   store <16 x i32> %s, <16 x i32> *%dst
   ret void
 }
 
 define void @vst4_v8i32(<8 x i32> *%src, <32 x i32> *%dst) {
 ; CHECK-LABEL: vst4_v8i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #48
-; CHECK-NEXT:    sub sp, #48
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vmov.f64 d2, d6
-; CHECK-NEXT:    vmov.32 r2, q5[0]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s5, s16
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[1]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s16, s13
-; CHECK-NEXT:    vmov.f32 s0, s13
-; CHECK-NEXT:    vmov.f32 s1, s17
-; CHECK-NEXT:    vmov.f64 d2, d7
-; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[2]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s5, s18
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s12, s15
-; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[3]
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s13, s19
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s14, s2
-; CHECK-NEXT:    vmov.32 r0, q2[0]
-; CHECK-NEXT:    vmov.f64 d10, d14
-; CHECK-NEXT:    vmov.f32 s15, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s21, s24
-; CHECK-NEXT:    vmov.32 r0, q2[1]
-; CHECK-NEXT:    vmov.f32 s22, s2
-; CHECK-NEXT:    vmov.f32 s23, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f64 d8, d15
-; CHECK-NEXT:    vmov.32 r0, q2[2]
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.32 r0, q2[3]
-; CHECK-NEXT:    vdup.32 q2, r0
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s0, s29
-; CHECK-NEXT:    vmov.f32 s17, s26
-; CHECK-NEXT:    vmov.f32 s24, s29
-; CHECK-NEXT:    vmov.f32 s1, s25
-; CHECK-NEXT:    vstrw.32 q0, [r1, #80]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s6
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s4, s31
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s5, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #96]
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vstrw.32 q1, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    add sp, #48
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i32>, <8 x i32>* %src, i32 0
   %l1 = load <8 x i32>, <8 x i32>* %s1, align 4
   %s2 = getelementptr <8 x i32>, <8 x i32>* %src, i32 1
   %l2 = load <8 x i32>, <8 x i32>* %s2, align 4
   %s3 = getelementptr <8 x i32>, <8 x i32>* %src, i32 2
   %l3 = load <8 x i32>, <8 x i32>* %s3, align 4
   %s4 = getelementptr <8 x i32>, <8 x i32>* %src, i32 3
   %l4 = load <8 x i32>, <8 x i32>* %s3, align 4
   %t1 = shufflevector <8 x i32> %l1, <8 x i32> %l2, <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 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x i32> %l3, <8 x i32> %l4, <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 12, i32 13, i32 14, i32 15>
   %s = shufflevector <16 x i32> %t1, <16 x i32> %t2, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
   store <32 x i32> %s, <32 x i32> *%dst
   ret void
 }
 
 define void @vst4_v16i32(<16 x i32> *%src, <64 x i32> *%dst) {
 ; CHECK-LABEL: vst4_v16i32:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #160
-; CHECK-NEXT:    sub sp, #160
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #176]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #112]
-; CHECK-NEXT:    vmov.32 r2, q1[2]
-; CHECK-NEXT:    vmov.f64 d8, d5
-; CHECK-NEXT:    vdup.32 q3, r2
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #160]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #128]
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #64]
-; CHECK-NEXT:    vstrw.32 q6, [sp, #144] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #96]
-; CHECK-NEXT:    vmov.f32 s17, s2
-; CHECK-NEXT:    vmov.f32 s18, s14
-; CHECK-NEXT:    vstrw.32 q6, [sp, #128] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s19, s15
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #32]
-; CHECK-NEXT:    vstrw.32 q3, [sp, #96] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vstrw.32 q6, [sp, #112] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q3, [sp, #80] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vstrw.32 q3, [sp, #64] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vstrw.32 q4, [r1, #224]
-; CHECK-NEXT:    vmov.f32 s16, s11
-; CHECK-NEXT:    vmov.32 r0, q1[3]
-; CHECK-NEXT:    vmov.f32 s17, s3
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #240]
-; CHECK-NEXT:    vmov.f64 d8, d4
-; CHECK-NEXT:    vmov.32 r0, q1[0]
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s17, s0
-; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #192]
-; CHECK-NEXT:    vmov.32 r0, q1[1]
-; CHECK-NEXT:    vmov.f32 s0, s9
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #144] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vldrw.u32 q6, [sp, #128] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q0, [r1, #208]
-; CHECK-NEXT:    vmov.f64 d0, d9
-; CHECK-NEXT:    vmov.32 r0, q2[2]
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s1, s26
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #160]
-; CHECK-NEXT:    vmov.f32 s0, s19
-; CHECK-NEXT:    vmov.32 r0, q2[3]
-; CHECK-NEXT:    vmov.f32 s1, s27
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s8, s15
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vmov.f64 d2, d6
-; CHECK-NEXT:    vstrw.32 q0, [r1, #176]
-; CHECK-NEXT:    vmov.32 r0, q5[0]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s5, s28
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vstrw.32 q1, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r0, q5[1]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s28, s13
-; CHECK-NEXT:    vmov.f32 s0, s13
-; CHECK-NEXT:    vmov.f32 s1, s29
-; CHECK-NEXT:    vmov.f64 d2, d7
-; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r0, q5[2]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.32 r0, q5[3]
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #96] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q5, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s5, s30
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s9, s31
-; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s10, s2
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.32 r0, q3[0]
-; CHECK-NEXT:    vmov.f64 d12, d4
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.32 r0, q3[1]
-; CHECK-NEXT:    vdup.32 q7, r0
-; CHECK-NEXT:    vmov.32 r0, q3[2]
-; CHECK-NEXT:    vmov.f32 s25, s20
-; CHECK-NEXT:    vmov.f32 s26, s2
-; CHECK-NEXT:    vmov.f64 d8, d5
-; CHECK-NEXT:    vmov.f32 s27, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s20, s9
-; CHECK-NEXT:    vmov.32 r0, q3[3]
-; CHECK-NEXT:    vmov.f32 s17, s22
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #128] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s2
-; CHECK-NEXT:    vmov.f32 s4, s11
-; CHECK-NEXT:    vmov.f32 s28, s9
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s19, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s5, s23
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f64 d0, d4
-; CHECK-NEXT:    vmov.f32 s1, s12
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #144] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s29, s21
-; CHECK-NEXT:    vmov.32 r0, q2[0]
-; CHECK-NEXT:    vdup.32 q5, r0
-; CHECK-NEXT:    vmov.32 r0, q2[1]
-; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vdup.32 q2, r0
-; CHECK-NEXT:    vmov.f32 s3, s23
-; CHECK-NEXT:    vstrw.32 q4, [r1, #96]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #128]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s14, s10
-; CHECK-NEXT:    vstrw.32 q1, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s15, s11
-; CHECK-NEXT:    vstrw.32 q6, [r1, #64]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q3, [r1, #144]
-; CHECK-NEXT:    vstrw.32 q7, [r1, #80]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    add sp, #160
+; CHECK-NEXT:    .pad #152
+; CHECK-NEXT:    sub sp, #152
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #176]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #128]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #64]
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #96]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #144]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    add.w r0, r1, #192
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    adds r1, #128
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vldmia r2, {d8, d9, d10, d11, d12, d13, d14, d15} @ 64-byte Reload
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    add sp, #152
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i32>, <16 x i32>* %src, i32 0
   %l1 = load <16 x i32>, <16 x i32>* %s1, align 4
   %s2 = getelementptr <16 x i32>, <16 x i32>* %src, i32 1
   %l2 = load <16 x i32>, <16 x i32>* %s2, align 4
   %s3 = getelementptr <16 x i32>, <16 x i32>* %src, i32 2
   %l3 = load <16 x i32>, <16 x i32>* %s3, align 4
entry:
   %s = shufflevector <8 x i16> %t1, <8 x i16> %t2, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
   store <16 x i16> %s, <16 x i16> *%dst
   ret void
 }
 
 define void @vst4_v8i16(<8 x i16> *%src, <32 x i16> *%dst) {
 ; CHECK-LABEL: vst4_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vmov.u16 r2, q1[2]
-; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q4[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q4[4], r2
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q4[5], r2
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.u16 r2, q4[0]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r2, q4[1]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r0, q5[2]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[3]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[4]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[5]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[6]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.u16 r0, q5[0]
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[1]
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[4]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[5]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vstrw.32 q4, [r1]
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q7[2], r0
-; CHECK-NEXT:    vmov.16 q7[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q7[6], r0
-; CHECK-NEXT:    vmov.16 q7[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[2]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[3]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[6]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vstrw.32 q5, [r1, #48]
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst41.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst42.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst43.16 {q0, q1, q2, q3}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i16>, <8 x i16>* %src, i32 0
   %l1 = load <8 x i16>, <8 x i16>* %s1, align 4
   %s2 = getelementptr <8 x i16>, <8 x i16>* %src, i32 1
   %l2 = load <8 x i16>, <8 x i16>* %s2, align 4
   %s3 = getelementptr <8 x i16>, <8 x i16>* %src, i32 2
   %l3 = load <8 x i16>, <8 x i16>* %s3, align 4
   %s4 = getelementptr <8 x i16>, <8 x i16>* %src, i32 3
   %l4 = load <8 x i16>, <8 x i16>* %s3, align 4
   %t1 = shufflevector <8 x i16> %l1, <8 x i16> %l2, <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 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x i16> %l3, <8 x i16> %l4, <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 12, i32 13, i32 14, i32 15>
   %s = shufflevector <16 x i16> %t1, <16 x i16> %t2, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
   store <32 x i16> %s, <32 x i16> *%dst
   ret void
 }
 
 define void @vst4_v16i16(<16 x i16> *%src, <64 x i16> *%dst) {
 ; CHECK-LABEL: vst4_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #64
-; CHECK-NEXT:    sub sp, #64
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #48]
-; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[0]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[1]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[1]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[0]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[1]
-; CHECK-NEXT:    vmov.16 q2[6], r2
-; CHECK-NEXT:    vmov.16 q2[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[5]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[2]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[2]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[3]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[3]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vstrw.32 q5, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[2]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[3]
-; CHECK-NEXT:    vmov.16 q2[6], r2
-; CHECK-NEXT:    vmov.16 q2[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[5]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[4]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[4]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[5]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[5]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vstrw.32 q5, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[4]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[5]
-; CHECK-NEXT:    vmov.16 q2[6], r2
-; CHECK-NEXT:    vmov.16 q2[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[5]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.16 q5[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[6]
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[6]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[7]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q4[7]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vstrw.32 q5, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.16 q7[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q7[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[6]
-; CHECK-NEXT:    vmov.16 q2[2], r2
-; CHECK-NEXT:    vmov.16 q2[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.16 q2[6], r2
-; CHECK-NEXT:    vmov.16 q2[7], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.16 q7[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.16 q7[3], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov.16 q7[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q0[5]
-; CHECK-NEXT:    vmov.16 q7[5], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.16 q7[6], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.16 q7[7], r2
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vstrw.32 q7, [r1, #48]
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.16 q0[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[0]
-; CHECK-NEXT:    vmov.16 q0[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.u16 r2, q6[1]
-; CHECK-NEXT:    vmov.16 q0[5], r2
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.u16 r2, q0[0]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vstrw.32 q2, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vstrw.32 q4, [r1, #80]
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q2[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vstrw.32 q2, [r1, #96]
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vstrw.32 q6, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    add sp, #64
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.16 {q0, q1, q2, q3}, [r0]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i16>, <16 x i16>* %src, i32 0
   %l1 = load <16 x i16>, <16 x i16>* %s1, align 4
   %s2 = getelementptr <16 x i16>, <16 x i16>* %src, i32 1
   %l2 = load <16 x i16>, <16 x i16>* %s2, align 4
   %s3 = getelementptr <16 x i16>, <16 x i16>* %src, i32 2
entry:
   %s = shufflevector <16 x i8> %t1, <16 x i8> %t2, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
   store <32 x i8> %s, <32 x i8> *%dst
   ret void
 }
 
 define void @vst4_v16i8(<16 x i8> *%src, <64 x i8> *%dst) {
 ; CHECK-LABEL: vst4_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vmov.u8 r2, q1[4]
-; CHECK-NEXT:    vmov.8 q4[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[4]
-; CHECK-NEXT:    vmov.8 q4[1], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[5]
-; CHECK-NEXT:    vmov.u8 r0, q3[4]
-; CHECK-NEXT:    vmov.8 q4[4], r2
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r2, q2[5]
-; CHECK-NEXT:    vmov.8 q4[5], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[6]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[5]
-; CHECK-NEXT:    vmov.8 q5[6], r0
-; CHECK-NEXT:    vmov.8 q4[8], r2
-; CHECK-NEXT:    vmov.u8 r2, q2[6]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.8 q4[9], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    vmov.u8 r0, q3[6]
-; CHECK-NEXT:    vmov.8 q4[12], r2
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r2, q2[7]
-; CHECK-NEXT:    vmov.8 q4[13], r2
-; CHECK-NEXT:    vmov.8 q5[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[7]
-; CHECK-NEXT:    vmov.u8 r2, q4[0]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.8 q0[0], r2
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vmov.u8 r2, q4[1]
-; CHECK-NEXT:    vmov.8 q0[1], r2
-; CHECK-NEXT:    vmov.u8 r0, q5[2]
-; CHECK-NEXT:    vmov.8 q0[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[3]
-; CHECK-NEXT:    vmov.8 q0[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[4]
-; CHECK-NEXT:    vmov.8 q0[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[5]
-; CHECK-NEXT:    vmov.8 q0[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[6]
-; CHECK-NEXT:    vmov.8 q0[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[7]
-; CHECK-NEXT:    vmov.8 q0[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[8]
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[9]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[10]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[11]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[12]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[13]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[14]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.u8 r0, q5[0]
-; CHECK-NEXT:    vmov.8 q4[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[1]
-; CHECK-NEXT:    vmov.8 q4[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[0]
-; CHECK-NEXT:    vmov.8 q6[2], r0
-; CHECK-NEXT:    vmov.8 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[1]
-; CHECK-NEXT:    vmov.8 q6[6], r0
-; CHECK-NEXT:    vmov.8 q6[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[2]
-; CHECK-NEXT:    vmov.8 q6[10], r0
-; CHECK-NEXT:    vmov.8 q6[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[3]
-; CHECK-NEXT:    vmov.8 q6[14], r0
-; CHECK-NEXT:    vmov.8 q6[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[2]
-; CHECK-NEXT:    vmov.8 q4[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[3]
-; CHECK-NEXT:    vmov.8 q4[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[4]
-; CHECK-NEXT:    vmov.8 q4[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[5]
-; CHECK-NEXT:    vmov.8 q4[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[6]
-; CHECK-NEXT:    vmov.8 q4[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[7]
-; CHECK-NEXT:    vmov.8 q4[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[8]
-; CHECK-NEXT:    vmov.8 q4[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[9]
-; CHECK-NEXT:    vmov.8 q4[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[10]
-; CHECK-NEXT:    vmov.8 q4[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[11]
-; CHECK-NEXT:    vmov.8 q4[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[12]
-; CHECK-NEXT:    vmov.8 q4[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[13]
-; CHECK-NEXT:    vmov.8 q4[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[14]
-; CHECK-NEXT:    vmov.8 q4[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[15]
-; CHECK-NEXT:    vmov.8 q4[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q6[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q6[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q6[13], r0
-; CHECK-NEXT:    vstrw.32 q4, [r1]
-; CHECK-NEXT:    vmov.u8 r0, q6[0]
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[1]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[12]
-; CHECK-NEXT:    vmov.8 q7[2], r0
-; CHECK-NEXT:    vmov.8 q7[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[13]
-; CHECK-NEXT:    vmov.8 q7[6], r0
-; CHECK-NEXT:    vmov.8 q7[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[14]
-; CHECK-NEXT:    vmov.8 q7[10], r0
-; CHECK-NEXT:    vmov.8 q7[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[15]
-; CHECK-NEXT:    vmov.8 q7[14], r0
-; CHECK-NEXT:    vmov.8 q7[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[2]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[3]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[4]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[5]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[6]
-; CHECK-NEXT:    vmov.8 q5[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[7]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[8]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[9]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[10]
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[11]
-; CHECK-NEXT:    vmov.8 q5[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[12]
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[13]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[14]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q7[15]
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q6[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q6[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q6[13], r0
-; CHECK-NEXT:    vstrw.32 q5, [r1, #48]
-; CHECK-NEXT:    vmov.u8 r0, q6[0]
-; CHECK-NEXT:    vmov.8 q1[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[1]
-; CHECK-NEXT:    vmov.8 q1[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[8]
-; CHECK-NEXT:    vmov.8 q2[2], r0
-; CHECK-NEXT:    vmov.8 q2[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[9]
-; CHECK-NEXT:    vmov.8 q2[6], r0
-; CHECK-NEXT:    vmov.8 q2[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[10]
-; CHECK-NEXT:    vmov.8 q2[10], r0
-; CHECK-NEXT:    vmov.8 q2[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[11]
-; CHECK-NEXT:    vmov.8 q2[14], r0
-; CHECK-NEXT:    vmov.8 q2[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q1[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q1[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[4]
-; CHECK-NEXT:    vmov.8 q1[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[5]
-; CHECK-NEXT:    vmov.8 q1[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q1[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q1[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[8]
-; CHECK-NEXT:    vmov.8 q1[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[9]
-; CHECK-NEXT:    vmov.8 q1[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q1[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q1[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[12]
-; CHECK-NEXT:    vmov.8 q1[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[13]
-; CHECK-NEXT:    vmov.8 q1[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q1[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q1[15], r0
-; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.8 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst41.8 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst42.8 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst43.8 {q0, q1, q2, q3}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i8>, <16 x i8>* %src, i32 0
   %l1 = load <16 x i8>, <16 x i8>* %s1, align 4
   %s2 = getelementptr <16 x i8>, <16 x i8>* %src, i32 1
   %l2 = load <16 x i8>, <16 x i8>* %s2, align 4
   %s3 = getelementptr <16 x i8>, <16 x i8>* %src, i32 2
   %l3 = load <16 x i8>, <16 x i8>* %s3, align 4
entry:
   %s = shufflevector <4 x float> %t1, <4 x float> %t2, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
   store <8 x float> %s, <8 x float> *%dst
   ret void
 }
 
 define void @vst4_v4f32(<4 x float> *%src, <16 x float> *%dst) {
 ; CHECK-LABEL: vst4_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s0, s9
-; CHECK-NEXT:    vmov.32 r0, q3[1]
-; CHECK-NEXT:    vdup.32 q4, r0
-; CHECK-NEXT:    vmov.f32 s1, s5
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vmov.32 r0, q3[0]
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vdup.32 q4, r0
-; CHECK-NEXT:    vmov.f32 s9, s4
-; CHECK-NEXT:    vmov.32 r0, q3[3]
-; CHECK-NEXT:    vmov.f32 s16, s8
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s20, s11
-; CHECK-NEXT:    vmov.32 r0, q3[2]
-; CHECK-NEXT:    vmov.f32 s8, s10
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s21, s7
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmov.f32 s9, s6
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vstrw.32 q4, [r1]
-; CHECK-NEXT:    vmov.f32 s10, s6
-; CHECK-NEXT:    vmov.f32 s23, s27
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vstrw.32 q5, [r1, #48]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <4 x float>, <4 x float>* %src, i32 0
   %l1 = load <4 x float>, <4 x float>* %s1, align 4
   %s2 = getelementptr <4 x float>, <4 x float>* %src, i32 1
   %l2 = load <4 x float>, <4 x float>* %s2, align 4
   %s3 = getelementptr <4 x float>, <4 x float>* %src, i32 2
   %l3 = load <4 x float>, <4 x float>* %s3, align 4
   %s4 = getelementptr <4 x float>, <4 x float>* %src, i32 3
   %l4 = load <4 x float>, <4 x float>* %s3, align 4
   %t1 = shufflevector <4 x float> %l1, <4 x float> %l2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %t2 = shufflevector <4 x float> %l3, <4 x float> %l4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %s = shufflevector <8 x float> %t1, <8 x float> %t2, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
   store <16 x float> %s, <16 x float> *%dst
   ret void
 }
 
 define void @vst4_v8f32(<8 x float> *%src, <32 x float> *%dst) {
 ; CHECK-LABEL: vst4_v8f32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #48
-; CHECK-NEXT:    sub sp, #48
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vmov.f64 d2, d6
-; CHECK-NEXT:    vmov.32 r2, q5[0]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s5, s16
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[1]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s16, s13
-; CHECK-NEXT:    vmov.f32 s0, s13
-; CHECK-NEXT:    vmov.f32 s1, s17
-; CHECK-NEXT:    vmov.f64 d2, d7
-; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[2]
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s5, s18
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f32 s12, s15
-; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r2, q5[3]
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s13, s19
-; CHECK-NEXT:    vdup.32 q0, r2
-; CHECK-NEXT:    vmov.f32 s14, s2
-; CHECK-NEXT:    vmov.32 r0, q2[0]
-; CHECK-NEXT:    vmov.f64 d10, d14
-; CHECK-NEXT:    vmov.f32 s15, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s21, s24
-; CHECK-NEXT:    vmov.32 r0, q2[1]
-; CHECK-NEXT:    vmov.f32 s22, s2
-; CHECK-NEXT:    vmov.f32 s23, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f64 d8, d15
-; CHECK-NEXT:    vmov.32 r0, q2[2]
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.32 r0, q2[3]
-; CHECK-NEXT:    vdup.32 q2, r0
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s0, s29
-; CHECK-NEXT:    vmov.f32 s17, s26
-; CHECK-NEXT:    vmov.f32 s24, s29
-; CHECK-NEXT:    vmov.f32 s1, s25
-; CHECK-NEXT:    vstrw.32 q0, [r1, #80]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s6
-; CHECK-NEXT:    vmov.f32 s19, s7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s4, s31
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s5, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #96]
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vstrw.32 q1, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    add sp, #48
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x float>, <8 x float>* %src, i32 0
   %l1 = load <8 x float>, <8 x float>* %s1, align 4
   %s2 = getelementptr <8 x float>, <8 x float>* %src, i32 1
   %l2 = load <8 x float>, <8 x float>* %s2, align 4
   %s3 = getelementptr <8 x float>, <8 x float>* %src, i32 2
   %l3 = load <8 x float>, <8 x float>* %s3, align 4
   %s4 = getelementptr <8 x float>, <8 x float>* %src, i32 3
   %l4 = load <8 x float>, <8 x float>* %s3, align 4
   %t1 = shufflevector <8 x float> %l1, <8 x float> %l2, <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 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x float> %l3, <8 x float> %l4, <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 12, i32 13, i32 14, i32 15>
   %s = shufflevector <16 x float> %t1, <16 x float> %t2, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
   store <32 x float> %s, <32 x float> *%dst
   ret void
 }
 
 define void @vst4_v16f32(<16 x float> *%src, <64 x float> *%dst) {
 ; CHECK-LABEL: vst4_v16f32:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5}
+; CHECK-NEXT:    push {r4, r5}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #160
-; CHECK-NEXT:    sub sp, #160
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #176]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #112]
-; CHECK-NEXT:    vmov.32 r2, q1[2]
-; CHECK-NEXT:    vmov.f64 d8, d5
-; CHECK-NEXT:    vdup.32 q3, r2
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #160]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #128]
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #64]
-; CHECK-NEXT:    vstrw.32 q6, [sp, #144] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #96]
-; CHECK-NEXT:    vmov.f32 s17, s2
-; CHECK-NEXT:    vmov.f32 s18, s14
-; CHECK-NEXT:    vstrw.32 q6, [sp, #128] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s19, s15
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #32]
-; CHECK-NEXT:    vstrw.32 q3, [sp, #96] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vstrw.32 q6, [sp, #112] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q3, [sp, #80] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vstrw.32 q3, [sp, #64] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vstrw.32 q4, [r1, #224]
-; CHECK-NEXT:    vmov.f32 s16, s11
-; CHECK-NEXT:    vmov.32 r0, q1[3]
-; CHECK-NEXT:    vmov.f32 s17, s3
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #240]
-; CHECK-NEXT:    vmov.f64 d8, d4
-; CHECK-NEXT:    vmov.32 r0, q1[0]
-; CHECK-NEXT:    vdup.32 q6, r0
-; CHECK-NEXT:    vmov.f32 s17, s0
-; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s27
-; CHECK-NEXT:    vstrw.32 q4, [r1, #192]
-; CHECK-NEXT:    vmov.32 r0, q1[1]
-; CHECK-NEXT:    vmov.f32 s0, s9
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #144] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vldrw.u32 q6, [sp, #128] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q0, [r1, #208]
-; CHECK-NEXT:    vmov.f64 d0, d9
-; CHECK-NEXT:    vmov.32 r0, q2[2]
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s1, s26
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vstrw.32 q0, [r1, #160]
-; CHECK-NEXT:    vmov.f32 s0, s19
-; CHECK-NEXT:    vmov.32 r0, q2[3]
-; CHECK-NEXT:    vmov.f32 s1, s27
-; CHECK-NEXT:    vdup.32 q1, r0
-; CHECK-NEXT:    vmov.f32 s8, s15
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vmov.f64 d2, d6
-; CHECK-NEXT:    vstrw.32 q0, [r1, #176]
-; CHECK-NEXT:    vmov.32 r0, q5[0]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s5, s28
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vstrw.32 q1, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r0, q5[1]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s28, s13
-; CHECK-NEXT:    vmov.f32 s0, s13
-; CHECK-NEXT:    vmov.f32 s1, s29
-; CHECK-NEXT:    vmov.f64 d2, d7
-; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.32 r0, q5[2]
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.32 r0, q5[3]
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #96] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q5, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s5, s30
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s9, s31
-; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s10, s2
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.32 r0, q3[0]
-; CHECK-NEXT:    vmov.f64 d12, d4
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.32 r0, q3[1]
-; CHECK-NEXT:    vdup.32 q7, r0
-; CHECK-NEXT:    vmov.32 r0, q3[2]
-; CHECK-NEXT:    vmov.f32 s25, s20
-; CHECK-NEXT:    vmov.f32 s26, s2
-; CHECK-NEXT:    vmov.f64 d8, d5
-; CHECK-NEXT:    vmov.f32 s27, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s20, s9
-; CHECK-NEXT:    vmov.32 r0, q3[3]
-; CHECK-NEXT:    vmov.f32 s17, s22
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #128] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s2
-; CHECK-NEXT:    vmov.f32 s4, s11
-; CHECK-NEXT:    vmov.f32 s28, s9
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s19, s3
-; CHECK-NEXT:    vdup.32 q0, r0
-; CHECK-NEXT:    vmov.f32 s5, s23
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vmov.f32 s7, s3
-; CHECK-NEXT:    vmov.f64 d0, d4
-; CHECK-NEXT:    vmov.f32 s1, s12
-; CHECK-NEXT:    vmov.f32 s12, s9
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #144] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s29, s21
-; CHECK-NEXT:    vmov.32 r0, q2[0]
-; CHECK-NEXT:    vdup.32 q5, r0
-; CHECK-NEXT:    vmov.32 r0, q2[1]
-; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vdup.32 q2, r0
-; CHECK-NEXT:    vmov.f32 s3, s23
-; CHECK-NEXT:    vstrw.32 q4, [r1, #96]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #128]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s14, s10
-; CHECK-NEXT:    vstrw.32 q1, [r1, #112]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s15, s11
-; CHECK-NEXT:    vstrw.32 q6, [r1, #64]
-; CHECK-NEXT:    vstrw.32 q0, [r1, #48]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q3, [r1, #144]
-; CHECK-NEXT:    vstrw.32 q7, [r1, #80]
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
-; CHECK-NEXT:    add sp, #160
+; CHECK-NEXT:    .pad #152
+; CHECK-NEXT:    sub sp, #152
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #176]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #128]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #64]
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #112]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #96]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vstmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vstmia r2, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Spill
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #144]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
+; CHECK-NEXT:    add r2, sp, #64
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    add.w r0, r1, #192
+; CHECK-NEXT:    vldmia sp, {d0, d1, d2, d3, d4, d5, d6, d7} @ 64-byte Reload
+; CHECK-NEXT:    adds r1, #128
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vldmia r2, {d8, d9, d10, d11, d12, d13, d14, d15} @ 64-byte Reload
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vst40.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.32 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.32 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    add sp, #152
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x float>, <16 x float>* %src, i32 0
   %l1 = load <16 x float>, <16 x float>* %s1, align 4
   %s2 = getelementptr <16 x float>, <16 x float>* %src, i32 1
   %l2 = load <16 x float>, <16 x float>* %s2, align 4
   %s3 = getelementptr <16 x float>, <16 x float>* %src, i32 2
   %l3 = load <16 x float>, <16 x float>* %s3, align 4
entry:
   %s = shufflevector <8 x half> %t1, <8 x half> %t2, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
   store <16 x half> %s, <16 x half> *%dst
   ret void
 }
 
 define void @vst4_v8f16(<8 x half> *%src, <32 x half> *%dst) {
 ; CHECK-LABEL: vst4_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8}
-; CHECK-NEXT:    vpush {d8}
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vmovx.f16 s16, s11
-; CHECK-NEXT:    vmov.16 q3[0], r2
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    vmov.16 q3[1], r3
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s3
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s7
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s2
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s6
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s16, s9
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s1
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s5
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s0, s0
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vstrw.32 q3, [r1, #16]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmovx.f16 s8, s8
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmovx.f16 s0, s4
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vstrw.32 q3, [r1]
-; CHECK-NEXT:    vpop {d8}
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vst40.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst41.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst42.16 {q0, q1, q2, q3}, [r1]
+; CHECK-NEXT:    vst43.16 {q0, q1, q2, q3}, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x half>, <8 x half>* %src, i32 0
   %l1 = load <8 x half>, <8 x half>* %s1, align 4
   %s2 = getelementptr <8 x half>, <8 x half>* %src, i32 1
   %l2 = load <8 x half>, <8 x half>* %s2, align 4
   %s3 = getelementptr <8 x half>, <8 x half>* %src, i32 2
   %l3 = load <8 x half>, <8 x half>* %s3, align 4
   %s4 = getelementptr <8 x half>, <8 x half>* %src, i32 3
   %l4 = load <8 x half>, <8 x half>* %s3, align 4
   %t1 = shufflevector <8 x half> %l1, <8 x half> %l2, <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 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x half> %l3, <8 x half> %l4, <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 12, i32 13, i32 14, i32 15>
   %s = shufflevector <16 x half> %t1, <16 x half> %t2, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
   store <32 x half> %s, <32 x half> *%dst
   ret void
 }
 
 define void @vst4_v16f16(<16 x half> *%src, <64 x half> *%dst) {
 ; CHECK-LABEL: vst4_v16f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14}
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmov r3, s22
-; CHECK-NEXT:    vmovx.f16 s0, s22
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.16 q6[0], r3
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov r2, s18
-; CHECK-NEXT:    vmov.16 q6[2], r2
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.16 q6[3], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmovx.f16 s0, s14
-; CHECK-NEXT:    vmov.16 q6[4], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmovx.f16 s0, s18
-; CHECK-NEXT:    vmov.16 q6[5], r2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov.16 q6[6], r2
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
-; CHECK-NEXT:    vmov.16 q6[7], r2
-; CHECK-NEXT:    vmov r0, s23
-; CHECK-NEXT:    vstrw.32 q6, [r1, #96]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    vmovx.f16 s28, s23
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov r0, s19
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmovx.f16 s28, s15
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmovx.f16 s28, s19
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmovx.f16 s28, s20
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vstrw.32 q6, [r1, #112]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmovx.f16 s28, s12
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmovx.f16 s28, s16
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov r0, s28
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmovx.f16 s20, s21
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov r0, s21
-; CHECK-NEXT:    vstrw.32 q6, [r1, #64]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.16 q6[1], r2
-; CHECK-NEXT:    vmov r0, s17
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmovx.f16 s12, s13
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s12, s17
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r2, s7
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vstrw.32 q6, [r1, #80]
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s6
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s2
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s16, s11
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s11
-; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s7
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s3
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s16, s8
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s4
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmovx.f16 s16, s0
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmovx.f16 s8, s9
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vstrw.32 q3, [r1]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.16 q3[1], r2
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmovx.f16 s4, s5
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmovx.f16 s0, s1
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vstrw.32 q3, [r1, #16]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    add.w r0, r1, #64
+; CHECK-NEXT:    vst40.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst41.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst42.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst43.16 {q4, q5, q6, q7}, [r1]
+; CHECK-NEXT:    vst40.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst41.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst42.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vst43.16 {q0, q1, q2, q3}, [r0]
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x half>, <16 x half>* %src, i32 0
   %l1 = load <16 x half>, <16 x half>* %s1, align 4
   %s2 = getelementptr <16 x half>, <16 x half>* %src, i32 1
   %l2 = load <16 x half>, <16 x half>* %s2, align 4
   %s3 = getelementptr <16 x half>, <16 x half>* %src, i32 2
   %l3 = load <16 x half>, <16 x half>* %s3, align 4

llvm/test/Transforms/InterleavedAccess/ARM/interleaved-accesses.ll

 ; CHECK-NEON-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld4.v4i32.p0i8(i8* [[TMP1]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 3
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 2
 ; CHECK-NEON-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 1
 ; CHECK-NEON-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_factor4(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
-; CHECK-MVE-NEXT:    [[V2:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
-; CHECK-MVE-NEXT:    [[V3:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.mve.vld4q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 3
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 2
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_factor4(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
 ; CHECK-NONE-NEXT:    [[V2:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
 ; CHECK-NONE-NEXT:    [[V3:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = load <16 x i32>, <16 x i32>* %ptr, align 4
   %v0 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %v1 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %v2 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %v3 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
   ret void
 }
 
 define void @store_factor2(<16 x i8>* %ptr, <8 x i8> %v0, <8 x i8> %v1) {
 ; CHECK-NEON-LABEL: @store_factor2(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i8> [[V0:%.*]], <8 x i8> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i8> [[V0]], <8 x i8> [[V1]], <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v8i8(i8* [[TMP1]], <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i8> [[V0:%.*]], <8 x i8> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i8> [[V0]], <8 x i8> [[V1]], <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v8i8(i8* [[TMP3]], <8 x i8> [[TMP1]], <8 x i8> [[TMP2]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor2(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i8> [[V0:%.*]], <8 x i8> [[V1:%.*]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
 ; CHECK-MVE-NEXT:    store <16 x i8> [[INTERLEAVED_VEC]], <16 x i8>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_factor2(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i8> [[V0:%.*]], <8 x i8> [[V1:%.*]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
 ; CHECK-NONE-NEXT:    store <16 x i8> [[INTERLEAVED_VEC]], <16 x i8>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <8 x i8> %v0, <8 x i8> %v1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x i8> %interleaved.vec, <16 x i8>* %ptr, align 4
   ret void
 }
 
 define void @store_factor3(<12 x i32>* %ptr, <4 x i32> %v0, <4 x i32> %v1, <4 x i32> %v2) {
 ; CHECK-NEON-LABEL: @store_factor3(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor3(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
;
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_factor4(<16 x i32>* %ptr, <4 x i32> %v0, <4 x i32> %v1, <4 x i32> %v2, <4 x i32> %v3) {
 ; CHECK-NEON-LABEL: @store_factor4(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor4(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
-; CHECK-MVE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 1)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 2)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 3)
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_factor4(
 ; CHECK-NONE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NONE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 1, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
 ; CHECK-NONE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
;
   %v3 = shufflevector <8 x i32*> %interleaved.vec, <8 x i32*> undef, <2 x i32> <i32 3, i32 7>
   ret void
 }
 
 define void @store_ptrvec_factor2(<4 x i32*>* %ptr, <2 x i32*> %v0, <2 x i32*> %v1) {
 ; CHECK-NEON-LABEL: @store_ptrvec_factor2(
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = ptrtoint <2 x i32*> [[V0:%.*]] to <2 x i32>
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = ptrtoint <2 x i32*> [[V1:%.*]] to <2 x i32>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <4 x i32*>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v2i32(i8* [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32*>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_ptrvec_factor2(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-MVE-NEXT:    store <4 x i32*> [[INTERLEAVED_VEC]], <4 x i32*>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_ptrvec_factor2(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NONE-NEXT:    store <4 x i32*> [[INTERLEAVED_VEC]], <4 x i32*>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <2 x i32*> %v0, <2 x i32*> %v1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   store <4 x i32*> %interleaved.vec, <4 x i32*>* %ptr, align 4
   ret void
 }
 
 define void @store_ptrvec_factor3(<6 x i32*>* %ptr, <2 x i32*> %v0, <2 x i32*> %v1, <2 x i32*> %v2) {
 ; CHECK-NEON-LABEL: @store_ptrvec_factor3(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <2 x i32*> [[V2:%.*]], <2 x i32*> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = ptrtoint <4 x i32*> [[S0]] to <4 x i32>
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = ptrtoint <4 x i32*> [[S1]] to <4 x i32>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <6 x i32*>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v2i32(i8* [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], <2 x i32> [[TMP6]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP6:%.*]] = bitcast <6 x i32*>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v2i32(i8* [[TMP6]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_ptrvec_factor3(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <2 x i32*> [[V2:%.*]], <2 x i32*> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i32*> [[S0]], <4 x i32*> [[S1]], <6 x i32> <i32 0, i32 2, i32 4, i32 1, i32 3, i32 5>
 ; CHECK-MVE-NEXT:    store <6 x i32*> [[INTERLEAVED_VEC]], <6 x i32*>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void

 }
 
 define void @store_ptrvec_factor4(<8 x i32*>* %ptr, <2 x i32*> %v0, <2 x i32*> %v1, <2 x i32*> %v2, <2 x i32*> %v3) {
 ; CHECK-NEON-LABEL: @store_ptrvec_factor4(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <2 x i32*> [[V2:%.*]], <2 x i32*> [[V3:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = ptrtoint <4 x i32*> [[S0]] to <4 x i32>
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = ptrtoint <4 x i32*> [[S1]] to <4 x i32>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <8 x i32*>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 6, i32 7>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], <2 x i32> [[TMP6]], <2 x i32> [[TMP7]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <2 x i32> <i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP7:%.*]] = bitcast <8 x i32*>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP7]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], <2 x i32> [[TMP6]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_ptrvec_factor4(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <2 x i32*> [[V0:%.*]], <2 x i32*> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <2 x i32*> [[V2:%.*]], <2 x i32*> [[V3:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i32*> [[S0]], <4 x i32*> [[S1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
 ; CHECK-MVE-NEXT:    store <8 x i32*> [[INTERLEAVED_VEC]], <8 x i32*>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void

 ; CHECK-NEON-LABEL: @load_undef_mask_factor2(
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
 ; CHECK-NEON-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32.p0i8(i8* [[TMP1]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 1
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_undef_mask_factor2(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <8 x i32> [[INTERLEAVED_VEC]], <8 x i32> undef, <4 x i32> <i32 undef, i32 2, i32 undef, i32 6>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <8 x i32> [[INTERLEAVED_VEC]], <8 x i32> undef, <4 x i32> <i32 undef, i32 3, i32 undef, i32 7>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_undef_mask_factor2(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <8 x i32> [[INTERLEAVED_VEC]], <8 x i32> undef, <4 x i32> <i32 undef, i32 2, i32 undef, i32 6>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <8 x i32> [[INTERLEAVED_VEC]], <8 x i32> undef, <4 x i32> <i32 undef, i32 3, i32 undef, i32 7>
 ; CHECK-NONE-NEXT:    ret void
 ;

 ; CHECK-NEON-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.neon.vld4.v4i32.p0i8(i8* [[TMP1]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 3
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 2
 ; CHECK-NEON-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 1
 ; CHECK-NEON-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_undef_mask_factor4(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 undef, i32 undef>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 undef, i32 undef>
-; CHECK-MVE-NEXT:    [[V2:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 undef, i32 undef>
-; CHECK-MVE-NEXT:    [[V3:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 undef, i32 undef>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.mve.vld4q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 3
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 2
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 0
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_undef_mask_factor4(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 undef, i32 undef>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 undef, i32 undef>
 ; CHECK-NONE-NEXT:    [[V2:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 undef, i32 undef>
 ; CHECK-NONE-NEXT:    [[V3:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 undef, i32 undef>
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = load <16 x i32>, <16 x i32>* %ptr, align 4
   %v0 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 undef, i32 undef>
   %v1 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 undef, i32 undef>
   %v2 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 undef, i32 undef>
   %v3 = shufflevector <16 x i32> %interleaved.vec, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 undef, i32 undef>
   ret void
 }
 
 define void @store_undef_mask_factor2(<8 x i32>* %ptr, <4 x i32> %v0, <4 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_undef_mask_factor2(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[V0]], <4 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[V0]], <4 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP3]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_undef_mask_factor2(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 2, i32 6, i32 3, i32 7>
-; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[V0]], <4 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP3]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP3]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], i32 1)
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_undef_mask_factor2(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 2, i32 6, i32 3, i32 7>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <4 x i32> %v0, <4 x i32> %v1, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 2, i32 6, i32 3, i32 7>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_undef_mask_factor3(<12 x i32>* %ptr, <4 x i32> %v0, <4 x i32> %v1, <4 x i32> %v2) {
 ; CHECK-NEON-LABEL: @store_undef_mask_factor3(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_undef_mask_factor3(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <12 x i32> <i32 0, i32 4, i32 undef, i32 1, i32 undef, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
;
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_undef_mask_factor4(<16 x i32>* %ptr, <4 x i32> %v0, <4 x i32> %v1, <4 x i32> %v2, <4 x i32> %v3) {
 ; CHECK-NEON-LABEL: @store_undef_mask_factor4(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_undef_mask_factor4(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <16 x i32> <i32 0, i32 4, i32 8, i32 undef, i32 undef, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
-; CHECK-MVE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 1)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 2)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP5]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 3)
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_undef_mask_factor4(
 ; CHECK-NONE-NEXT:    [[S0:%.*]] = shufflevector <4 x i32> [[V0:%.*]], <4 x i32> [[V1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NONE-NEXT:    [[S1:%.*]] = shufflevector <4 x i32> [[V2:%.*]], <4 x i32> [[V3:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[S0]], <8 x i32> [[S1]], <16 x i32> <i32 0, i32 4, i32 8, i32 undef, i32 undef, i32 5, i32 9, i32 13, i32 2, i32 6, i32 10, i32 14, i32 3, i32 7, i32 11, i32 15>
 ; CHECK-NONE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
;
   %v0 = shufflevector <8 x i32> %interleaved.vec, <8 x i32> undef, <2 x i32> <i32 0, i32 3>
   %v1 = shufflevector <8 x i32> %interleaved.vec, <8 x i32> undef, <2 x i32> <i32 1, i32 4>
   %v2 = shufflevector <8 x i32> %interleaved.vec, <8 x i32> undef, <2 x i32> <i32 2, i32 5>
   ret void
 }
 
 define void @store_address_space(<4 x i32> addrspace(1)* %ptr, <2 x i32> %v0, <2 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_address_space(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> addrspace(1)* [[PTR:%.*]] to i8 addrspace(1)*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> [[V0:%.*]], <2 x i32> [[V1:%.*]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i32> [[V0]], <2 x i32> [[V1]], <2 x i32> <i32 2, i32 3>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p1i8.v2i32(i8 addrspace(1)* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], i32 0)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[V0:%.*]], <2 x i32> [[V1:%.*]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> [[V0]], <2 x i32> [[V1]], <2 x i32> <i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = bitcast <4 x i32> addrspace(1)* [[PTR:%.*]] to i8 addrspace(1)*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p1i8.v2i32(i8 addrspace(1)* [[TMP3]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], i32 0)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_address_space(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <2 x i32> [[V0:%.*]], <2 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-MVE-NEXT:    store <4 x i32> [[INTERLEAVED_VEC]], <4 x i32> addrspace(1)* [[PTR:%.*]]
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_address_space(

 ;
   %interleaved.vec = shufflevector <3 x float> %v0, <3 x float> undef, <3 x i32> <i32 0, i32 2, i32 undef>
   store <3 x float> %interleaved.vec, <3 x float>* %ptr, align 16
   ret void
 }
 
 define void @store_general_mask_factor4(<8 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor4(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor4(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
 ; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor4(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor4_undefbeg(<8 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor4_undefbeg(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor4_undefbeg(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 undef, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
 ; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor4_undefbeg(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 undef, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <8 x i32> <i32 undef, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 9>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor4_undefend(<8 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor4_undefend(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor4_undefend(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 undef>
 ; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor4_undefend(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 undef>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <8 x i32> <i32 4, i32 16, i32 32, i32 8, i32 5, i32 17, i32 33, i32 undef>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor4_undefmid(<8 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor4_undefmid(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 16, i32 17>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 32, i32 33>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor4_undefmid(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 undef, i32 32, i32 8, i32 5, i32 17, i32 undef, i32 9>
 ; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor4_undefmid(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 undef, i32 32, i32 8, i32 5, i32 17, i32 undef, i32 9>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <8 x i32> <i32 4, i32 undef, i32 32, i32 8, i32 5, i32 17, i32 undef, i32 9>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor4_undefmulti(<8 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor4_undefmulti(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <2 x i32> <i32 4, i32 5>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 0, i32 1>
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 0, i32 1>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <2 x i32> <i32 8, i32 9>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast <8 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v2i32(i8* [[TMP5]], <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor4_undefmulti(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 undef, i32 undef, i32 8, i32 undef, i32 undef, i32 undef, i32 9>
 ; CHECK-MVE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor4_undefmulti(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <8 x i32> <i32 4, i32 undef, i32 undef, i32 8, i32 undef, i32 undef, i32 undef, i32 9>
 ; CHECK-NONE-NEXT:    store <8 x i32> [[INTERLEAVED_VEC]], <8 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <8 x i32> <i32 4, i32 undef, i32 undef, i32 8, i32 undef, i32 undef, i32 undef, i32 9>
   store <8 x i32> %interleaved.vec, <8 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor3(<12 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor3(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor3(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 4, i32 32, i32 16, i32 5, i32 33, i32 17, i32 6, i32 34, i32 18, i32 7, i32 35, i32 19>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor3(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 4, i32 32, i32 16, i32 5, i32 33, i32 17, i32 6, i32 34, i32 18, i32 7, i32 35, i32 19>
 ; CHECK-NONE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <12 x i32> <i32 4, i32 32, i32 16, i32 5, i32 33, i32 17, i32 6, i32 34, i32 18, i32 7, i32 35, i32 19>
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor3_undefmultimid(<12 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor3_undefmultimid(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor3_undefmultimid(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 4, i32 32, i32 16, i32 undef, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 7, i32 35, i32 19>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor3_undefmultimid(

 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <12 x i32> <i32 4, i32 32, i32 16, i32 undef, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 8, i32 35, i32 19>
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor3_undeflane(<12 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor3_undeflane(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor3_undeflane(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 undef, i32 32, i32 16, i32 undef, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 undef, i32 35, i32 19>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor3_undeflane(

 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <12 x i32> <i32 undef, i32 32, i32 16, i32 undef, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 2, i32 35, i32 19>
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor3_endstart_pass(<12 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor3_endstart_pass(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor3_endstart_pass(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 undef, i32 32, i32 16, i32 undef, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 7, i32 35, i32 19>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor3_endstart_pass(

 ;
   %interleaved.vec = shufflevector <32 x i32> %v0, <32 x i32> %v1, <12 x i32> <i32 undef, i32 32, i32 16, i32 0, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 undef, i32 35, i32 19>
   store <12 x i32> %interleaved.vec, <12 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_general_mask_factor3_midstart_pass(<12 x i32>* %ptr, <32 x i32> %v0, <32 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_general_mask_factor3_midstart_pass(
-; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP1:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <32 x i32> [[V0]], <32 x i32> [[V1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast <12 x i32>* [[PTR:%.*]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_general_mask_factor3_midstart_pass(
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <32 x i32> [[V0:%.*]], <32 x i32> [[V1:%.*]], <12 x i32> <i32 undef, i32 32, i32 16, i32 1, i32 33, i32 17, i32 undef, i32 34, i32 18, i32 undef, i32 35, i32 19>
 ; CHECK-MVE-NEXT:    store <12 x i32> [[INTERLEAVED_VEC]], <12 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_general_mask_factor3_midstart_pass(

 ; CHECK-NEON-NEXT:    [[VLDN1:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32.p0i8(i8* [[TMP6]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 1
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 0
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP7]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP8]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_factor2_wide2(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 0
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = getelementptr i32, i32* [[TMP1]], i32 8
+; CHECK-MVE-NEXT:    [[VLDN1:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP4]])
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 1
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 0
+; CHECK-MVE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_factor2_wide2(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32> [[INTERLEAVED_VEC]], <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
 ; CHECK-NONE-NEXT:    ret void
 ;

 ; CHECK-NEON-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP11]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-NEON-NEXT:    [[TMP15:%.*]] = shufflevector <8 x i32> [[TMP13]], <8 x i32> [[TMP14]], <12 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>
 ; CHECK-NEON-NEXT:    [[TMP16:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP8]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP17:%.*]] = shufflevector <4 x i32> [[TMP12]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-NEON-NEXT:    [[TMP18:%.*]] = shufflevector <8 x i32> [[TMP16]], <8 x i32> [[TMP17]], <12 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>
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_factor2_wide3(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <24 x i32>, <24 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <24 x i32> [[INTERLEAVED_VEC]], <24 x i32> undef, <12 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <24 x i32> [[INTERLEAVED_VEC]], <24 x i32> undef, <12 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <24 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 0
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = getelementptr i32, i32* [[TMP1]], i32 8
+; CHECK-MVE-NEXT:    [[VLDN1:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP4]])
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 1
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 0
+; CHECK-MVE-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP4]], i32 8
+; CHECK-MVE-NEXT:    [[VLDN2:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP7]])
+; CHECK-MVE-NEXT:    [[TMP8:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN2]], 1
+; CHECK-MVE-NEXT:    [[TMP9:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN2]], 0
+; CHECK-MVE-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i32> [[TMP8]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-MVE-NEXT:    [[TMP12:%.*]] = shufflevector <8 x i32> [[TMP10]], <8 x i32> [[TMP11]], <12 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>
+; CHECK-MVE-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP9]], <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-MVE-NEXT:    [[TMP15:%.*]] = shufflevector <8 x i32> [[TMP13]], <8 x i32> [[TMP14]], <12 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>
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_factor2_wide3(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <24 x i32>, <24 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <24 x i32> [[INTERLEAVED_VEC]], <24 x i32> undef, <12 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <24 x i32> [[INTERLEAVED_VEC]], <24 x i32> undef, <12 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23>
 ; CHECK-NONE-NEXT:    ret void
 ;

 ; CHECK-NEON-NEXT:    [[TMP12:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN1]], 0
 ; CHECK-NEON-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP9]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP10]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP15:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> [[TMP11]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP16:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_factor4_wide(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <32 x i32>, <32 x i32>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
-; CHECK-MVE-NEXT:    [[V2:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
-; CHECK-MVE-NEXT:    [[V3:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <32 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.mve.vld4q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 3
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 2
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN]], 0
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[TMP1]], i32 16
+; CHECK-MVE-NEXT:    [[VLDN1:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.arm.mve.vld4q.v4i32.p0i32(i32* [[TMP6]])
+; CHECK-MVE-NEXT:    [[TMP7:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN1]], 3
+; CHECK-MVE-NEXT:    [[TMP8:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN1]], 2
+; CHECK-MVE-NEXT:    [[TMP9:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN1]], 1
+; CHECK-MVE-NEXT:    [[TMP10:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLDN1]], 0
+; CHECK-MVE-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP7]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP8]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP9]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> [[TMP10]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_factor4_wide(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <32 x i32>, <32 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
 ; CHECK-NONE-NEXT:    [[V2:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
 ; CHECK-NONE-NEXT:    [[V3:%.*]] = shufflevector <32 x i32> [[INTERLEAVED_VEC]], <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = load <32 x i32>, <32 x i32>* %ptr, align 4
   %v0 = shufflevector <32 x i32> %interleaved.vec, <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   %v1 = shufflevector <32 x i32> %interleaved.vec, <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
   %v2 = shufflevector <32 x i32> %interleaved.vec, <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
   %v3 = shufflevector <32 x i32> %interleaved.vec, <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
   ret void
 }
 
 define void @store_factor2_wide(<16 x i32>* %ptr, <8 x i32> %v0, <8 x i32> %v1) {
 ; CHECK-NEON-LABEL: @store_factor2_wide(
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to i8*
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP1]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP4]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr i32, i32* [[TMP1]], i32 8
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to i8*
-; CHECK-NEON-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP8:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP5]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst2.p0i8.v4i32(i8* [[TMP8]], <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor2_wide(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-MVE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], i32 1)
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = getelementptr i32, i32* [[TMP1]], i32 8
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i32> [[V0]], <8 x i32> [[V1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP4]], <4 x i32> [[TMP5]], <4 x i32> [[TMP6]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst2q.p0i32.v4i32(i32* [[TMP4]], <4 x i32> [[TMP5]], <4 x i32> [[TMP6]], i32 1)
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_factor2_wide(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
 ; CHECK-NONE-NEXT:    store <16 x i32> [[INTERLEAVED_VEC]], <16 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void
 ;
   %interleaved.vec = shufflevector <8 x i32> %v0, <8 x i32> %v1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   store <16 x i32> %interleaved.vec, <16 x i32>* %ptr, align 4
   ret void
 }
 
 define void @store_factor3_wide(<24 x i32>* %ptr, <8 x i32> %v0, <8 x i32> %v1, <8 x i32> %v2) {
 ; CHECK-NEON-LABEL: @store_factor3_wide(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <8 x i32> [[V2:%.*]], <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <24 x i32>* [[PTR:%.*]] to i32*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to i8*
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP1]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP5]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[TMP1]], i32 12
-; CHECK-NEON-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to i8*
-; CHECK-NEON-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    [[TMP10:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; CHECK-NEON-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP6]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst3.p0i8.v4i32(i8* [[TMP10]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor3_wide(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <8 x i32> [[V2:%.*]], <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
 ; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>
 ; CHECK-MVE-NEXT:    store <24 x i32> [[INTERLEAVED_VEC]], <24 x i32>* [[PTR:%.*]], align 4
 ; CHECK-MVE-NEXT:    ret void
;
   ret void
 }
 
 define void @store_factor4_wide(<32 x i32>* %ptr, <8 x i32> %v0, <8 x i32> %v1, <8 x i32> %v2, <8 x i32> %v3) {
 ; CHECK-NEON-LABEL: @store_factor4_wide(
 ; CHECK-NEON-NEXT:    [[S0:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-NEON-NEXT:    [[S1:%.*]] = shufflevector <8 x i32> [[V2:%.*]], <8 x i32> [[V3:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = bitcast <32 x i32>* [[PTR:%.*]] to i32*
-; CHECK-NEON-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to i8*
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], <4 x i32> [[TMP6]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP2:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEON-NEXT:    [[TMP3:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEON-NEXT:    [[TMP4:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-NEON-NEXT:    [[TMP5:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; CHECK-NEON-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP1]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP6]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 4)
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP1]], i32 16
-; CHECK-NEON-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to i8*
-; CHECK-NEON-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEON-NEXT:    [[TMP10:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEON-NEXT:    [[TMP11:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; CHECK-NEON-NEXT:    [[TMP12:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], i32 4)
+; CHECK-NEON-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEON-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEON-NEXT:    [[TMP10:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; CHECK-NEON-NEXT:    [[TMP11:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; CHECK-NEON-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP7]] to i8*
+; CHECK-NEON-NEXT:    call void @llvm.arm.neon.vst4.p0i8.v4i32(i8* [[TMP12]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i32 4)
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @store_factor4_wide(
 ; CHECK-MVE-NEXT:    [[S0:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-MVE-NEXT:    [[S1:%.*]] = shufflevector <8 x i32> [[V2:%.*]], <8 x i32> [[V3:%.*]], <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 12, i32 13, i32 14, i32 15>
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
-; CHECK-MVE-NEXT:    store <32 x i32> [[INTERLEAVED_VEC]], <32 x i32>* [[PTR:%.*]], align 4
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <32 x i32>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 1)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 2)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP1]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], i32 3)
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[TMP1]], i32 16
+; CHECK-MVE-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; CHECK-MVE-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; CHECK-MVE-NEXT:    [[TMP10:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], i32 0)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], i32 1)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], i32 2)
+; CHECK-MVE-NEXT:    call void @llvm.arm.mve.vst4q.p0i32.v4i32(i32* [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], i32 3)
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @store_factor4_wide(
 ; CHECK-NONE-NEXT:    [[S0:%.*]] = shufflevector <8 x i32> [[V0:%.*]], <8 x i32> [[V1:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-NONE-NEXT:    [[S1:%.*]] = shufflevector <8 x i32> [[V2:%.*]], <8 x i32> [[V3:%.*]], <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 12, i32 13, i32 14, i32 15>
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i32> [[S0]], <16 x i32> [[S1]], <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 1, i32 9, i32 17, i32 25, i32 2, i32 10, i32 18, i32 26, i32 3, i32 11, i32 19, i32 27, i32 4, i32 12, i32 20, i32 28, i32 5, i32 13, i32 21, i32 29, i32 6, i32 14, i32 22, i32 30, i32 7, i32 15, i32 23, i32 31>
 ; CHECK-NONE-NEXT:    store <32 x i32> [[INTERLEAVED_VEC]], <32 x i32>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    ret void

 ; CHECK-NEON-NEXT:    [[TMP10:%.*]] = inttoptr <4 x i32> [[TMP9]] to <4 x i32*>
 ; CHECK-NEON-NEXT:    [[TMP11:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 0
 ; CHECK-NEON-NEXT:    [[TMP12:%.*]] = inttoptr <4 x i32> [[TMP11]] to <4 x i32*>
 ; CHECK-NEON-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32*> [[TMP4]], <4 x i32*> [[TMP10]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32*> [[TMP6]], <4 x i32*> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEON-NEXT:    ret void
 ;
 ; CHECK-MVE-LABEL: @load_factor2_wide_pointer(
-; CHECK-MVE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32*>, <16 x i32*>* [[PTR:%.*]], align 4
-; CHECK-MVE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32*> [[INTERLEAVED_VEC]], <16 x i32*> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-MVE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32*> [[INTERLEAVED_VEC]], <16 x i32*> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-MVE-NEXT:    [[TMP1:%.*]] = bitcast <16 x i32*>* [[PTR:%.*]] to i32*
+; CHECK-MVE-NEXT:    [[VLDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP1]])
+; CHECK-MVE-NEXT:    [[TMP2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 1
+; CHECK-MVE-NEXT:    [[TMP3:%.*]] = inttoptr <4 x i32> [[TMP2]] to <4 x i32*>
+; CHECK-MVE-NEXT:    [[TMP4:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN]], 0
+; CHECK-MVE-NEXT:    [[TMP5:%.*]] = inttoptr <4 x i32> [[TMP4]] to <4 x i32*>
+; CHECK-MVE-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[TMP1]], i32 8
+; CHECK-MVE-NEXT:    [[VLDN1:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.arm.mve.vld2q.v4i32.p0i32(i32* [[TMP6]])
+; CHECK-MVE-NEXT:    [[TMP7:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 1
+; CHECK-MVE-NEXT:    [[TMP8:%.*]] = inttoptr <4 x i32> [[TMP7]] to <4 x i32*>
+; CHECK-MVE-NEXT:    [[TMP9:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLDN1]], 0
+; CHECK-MVE-NEXT:    [[TMP10:%.*]] = inttoptr <4 x i32> [[TMP9]] to <4 x i32*>
+; CHECK-MVE-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i32*> [[TMP3]], <4 x i32*> [[TMP8]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-MVE-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32*> [[TMP5]], <4 x i32*> [[TMP10]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-MVE-NEXT:    ret void
 ;
 ; CHECK-NONE-LABEL: @load_factor2_wide_pointer(
 ; CHECK-NONE-NEXT:    [[INTERLEAVED_VEC:%.*]] = load <16 x i32*>, <16 x i32*>* [[PTR:%.*]], align 4
 ; CHECK-NONE-NEXT:    [[V0:%.*]] = shufflevector <16 x i32*> [[INTERLEAVED_VEC]], <16 x i32*> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
 ; CHECK-NONE-NEXT:    [[V1:%.*]] = shufflevector <16 x i32*> [[INTERLEAVED_VEC]], <16 x i32*> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
 ; CHECK-NONE-NEXT:    ret void
 ;

llvm/test/Transforms/LoopVectorize/ARM/mve-interleaved-cost.ll

entry:
   br label %for.body
 
 ; VF_2-LABEL:  Checking a loop in "i8_factor_2"
 ; VF_2:          Found an estimated cost of 20 for VF 2 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i8 0, i8* %tmp0, align 1
 ; VF_2-NEXT:     Found an estimated cost of 12 for VF 2 For instruction: store i8 0, i8* %tmp1, align 1
 ; VF_4-LABEL:  Checking a loop in "i8_factor_2"
-; VF_4:          Found an estimated cost of 72 for VF 4 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
+; VF_4:          Found an estimated cost of 4 for VF 4 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i8 0, i8* %tmp0, align 1
-; VF_4-NEXT:     Found an estimated cost of 40 for VF 4 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_4-NEXT:     Found an estimated cost of 4 for VF 4 For instruction: store i8 0, i8* %tmp1, align 1
 ; VF_8-LABEL:  Checking a loop in "i8_factor_2"
-; VF_8:          Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
+; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp0, align 1
-; VF_8-NEXT:     Found an estimated cost of 2 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
 ; VF_16-LABEL: Checking a loop in "i8_factor_2"
-; VF_16:         Found an estimated cost of 2 for VF 16 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
+; VF_16:         Found an estimated cost of 4 for VF 16 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp0, align 1
-; VF_16-NEXT:    Found an estimated cost of 2 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 4 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i8.2, %i8.2* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i8.2, %i8.2* %data, i64 %i, i32 1
   %tmp2 = load i8, i8* %tmp0, align 1
   %tmp3 = load i8, i8* %tmp1, align 1
   store i8 0, i8* %tmp0, align 1
   store i8 0, i8* %tmp1, align 1
entry:
   br label %for.body
 
 ; VF_2-LABEL:  Checking a loop in "i16_factor_2"
 ; VF_2:          Found an estimated cost of 20 for VF 2 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i16 0, i16* %tmp0, align 2
 ; VF_2-NEXT:     Found an estimated cost of 12 for VF 2 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_4-LABEL:  Checking a loop in "i16_factor_2"
-; VF_4:          Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_4:          Found an estimated cost of 4 for VF 4 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp0, align 2
-; VF_4-NEXT:     Found an estimated cost of 2 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 4 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_8-LABEL:  Checking a loop in "i16_factor_2"
-; VF_8:          Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp0, align 2
-; VF_8-NEXT:     Found an estimated cost of 2 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_16-LABEL: Checking a loop in "i16_factor_2"
-; VF_16:         Found an estimated cost of 4 for VF 16 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp0, align 2
-; VF_16-NEXT:    Found an estimated cost of 4 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i16.2, %i16.2* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i16.2, %i16.2* %data, i64 %i, i32 1
   %tmp2 = load i16, i16* %tmp0, align 2
   %tmp3 = load i16, i16* %tmp1, align 2
   store i16 0, i16* %tmp0, align 2
   store i16 0, i16* %tmp1, align 2
   %i.next = add nuw nsw i64 %i, 1
   %cond = icmp slt i64 %i.next, %n
   br i1 %cond, label %for.body, label %for.end
 
 for.end:
   ret void
 }
 
 %i32.2 = type {i32, i32}
 define void @i32_factor_2(%i32.2* %data, i64 %n) #0 {
 entry:
   br label %for.body
 
 ; VF_2-LABEL:  Checking a loop in "i32_factor_2"
-; VF_2:          Found an estimated cost of 2 for VF 2 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_2:          Found an estimated cost of 20 for VF 2 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp0, align 4
-; VF_2-NEXT:     Found an estimated cost of 2 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 12 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_4-LABEL:  Checking a loop in "i32_factor_2"
-; VF_4:          Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_4:          Found an estimated cost of 4 for VF 4 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp0, align 4
-; VF_4-NEXT:     Found an estimated cost of 2 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 4 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_8-LABEL:  Checking a loop in "i32_factor_2"
-; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_8:          Found an estimated cost of 8 for VF 8 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp0, align 4
-; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 8 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_16-LABEL: Checking a loop in "i32_factor_2"
-; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_16:         Found an estimated cost of 16 for VF 16 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp0, align 4
-; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 16 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i32.2, %i32.2* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i32.2, %i32.2* %data, i64 %i, i32 1
   %tmp2 = load i32, i32* %tmp0, align 4
   %tmp3 = load i32, i32* %tmp1, align 4
   store i32 0, i32* %tmp0, align 4
   store i32 0, i32* %tmp1, align 4

 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store half 0xH0000, half* %tmp0, align 2
 ; VF_2-NEXT:     Found an estimated cost of 12 for VF 2 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_4-LABEL:  Checking a loop in "f16_factor_2"
 ; VF_4:          Found an estimated cost of 72 for VF 4 For instruction: %tmp2 = load half, half* %tmp0, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load half, half* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store half 0xH0000, half* %tmp0, align 2
 ; VF_4-NEXT:     Found an estimated cost of 40 for VF 4 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_8-LABEL:  Checking a loop in "f16_factor_2"
-; VF_8:          Found an estimated cost of 272 for VF 8 For instruction: %tmp2 = load half, half* %tmp0, align 2
+; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load half, half* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load half, half* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store half 0xH0000, half* %tmp0, align 2
-; VF_8-NEXT:     Found an estimated cost of 144 for VF 8 For instruction: store half 0xH0000, half* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_16-LABEL: Checking a loop in "f16_factor_2"
-; VF_16:         Found an estimated cost of 1056 for VF 16 For instruction: %tmp2 = load half, half* %tmp0, align 2
+; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load half, half* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load half, half* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store half 0xH0000, half* %tmp0, align 2
-; VF_16-NEXT:    Found an estimated cost of 544 for VF 16 For instruction: store half 0xH0000, half* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store half 0xH0000, half* %tmp1, align 2
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %f16.2, %f16.2* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %f16.2, %f16.2* %data, i64 %i, i32 1
   %tmp2 = load half, half* %tmp0, align 2
   %tmp3 = load half, half* %tmp1, align 2
   store half 0.0, half* %tmp0, align 2
   store half 0.0, half* %tmp1, align 2
   %i.next = add nuw nsw i64 %i, 1
   %cond = icmp slt i64 %i.next, %n
   br i1 %cond, label %for.body, label %for.end
 
 for.end:
   ret void
 }
 
 %f32.2 = type {float, float}
 define void @f32_factor_2(%f32.2* %data, i64 %n) #0 {
 entry:
   br label %for.body
 
 ; VF_2-LABEL:  Checking a loop in "f32_factor_2"
-; VF_2:          Found an estimated cost of 2 for VF 2 For instruction: %tmp2 = load float, float* %tmp0, align 4
+; VF_2:          Found an estimated cost of 20 for VF 2 For instruction: %tmp2 = load float, float* %tmp0, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load float, float* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store float 0.000000e+00, float* %tmp0, align 4
-; VF_2-NEXT:     Found an estimated cost of 2 for VF 2 For instruction: store float 0.000000e+00, float* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 12 for VF 2 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_4-LABEL:  Checking a loop in "f32_factor_2"
-; VF_4:          Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load float, float* %tmp0, align 4
+; VF_4:          Found an estimated cost of 4 for VF 4 For instruction: %tmp2 = load float, float* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load float, float* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store float 0.000000e+00, float* %tmp0, align 4
-; VF_4-NEXT:     Found an estimated cost of 2 for VF 4 For instruction: store float 0.000000e+00, float* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 4 for VF 4 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_8-LABEL:  Checking a loop in "f32_factor_2"
-; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load float, float* %tmp0, align 4
+; VF_8:          Found an estimated cost of 8 for VF 8 For instruction: %tmp2 = load float, float* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load float, float* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store float 0.000000e+00, float* %tmp0, align 4
-; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store float 0.000000e+00, float* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 8 for VF 8 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_16-LABEL: Checking a loop in "f32_factor_2"
-; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load float, float* %tmp0, align 4
+; VF_16:         Found an estimated cost of 16 for VF 16 For instruction: %tmp2 = load float, float* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load float, float* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store float 0.000000e+00, float* %tmp0, align 4
-; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store float 0.000000e+00, float* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 16 for VF 16 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %f32.2, %f32.2* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %f32.2, %f32.2* %data, i64 %i, i32 1
   %tmp2 = load float, float* %tmp0, align 4
   %tmp3 = load float, float* %tmp1, align 4
   store float 0.0, float* %tmp0, align 4
   store float 0.0, float* %tmp1, align 4

 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i8, i8* %tmp1, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i8, i8* %tmp2, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i8, i8* %tmp3, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp0, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp2, align 1
 ; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i8 0, i8* %tmp3, align 1
 ; VF_16-LABEL: Checking a loop in "i8_factor_4"
-; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i8, i8* %tmp0, align 1
+; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp4 = load i8, i8* %tmp0, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i8, i8* %tmp1, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i8, i8* %tmp2, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i8, i8* %tmp3, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp0, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp2, align 1
-; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i8 0, i8* %tmp3, align 1
+; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store i8 0, i8* %tmp3, align 1
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 1
   %tmp2 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 2
   %tmp3 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 3
   %tmp4 = load i8, i8* %tmp0, align 1
   %tmp5 = load i8, i8* %tmp1, align 1

 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp0, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp2, align 2
 ; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store i16 0, i16* %tmp3, align 2
 ; VF_8-LABEL:  Checking a loop in "i16_factor_4"
-; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
+; VF_8:          Found an estimated cost of 8 for VF 8 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp2, align 2
-; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i16 0, i16* %tmp3, align 2
+; VF_8-NEXT:     Found an estimated cost of 8 for VF 8 For instruction: store i16 0, i16* %tmp3, align 2
 ; VF_16-LABEL: Checking a loop in "i16_factor_4"
-; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
+; VF_16:         Found an estimated cost of 16 for VF 16 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp2, align 2
-; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i16 0, i16* %tmp3, align 2
+; VF_16-NEXT:    Found an estimated cost of 16 for VF 16 For instruction: store i16 0, i16* %tmp3, align 2
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 1
   %tmp2 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 2
   %tmp3 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 3
   %tmp4 = load i16, i16* %tmp0, align 2
   %tmp5 = load i16, i16* %tmp1, align 2

 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp0, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp2, align 4
 ; VF_2-NEXT:     Found an estimated cost of 24 for VF 2 For instruction: store i32 0, i32* %tmp3, align 4
 ; VF_4-LABEL:  Checking a loop in "i32_factor_4"
-; VF_4:          Found an estimated cost of 144 for VF 4 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_4:          Found an estimated cost of 8 for VF 4 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp2, align 4
-; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_4-NEXT:     Found an estimated cost of 8 for VF 4 For instruction: store i32 0, i32* %tmp3, align 4
 ; VF_8-LABEL:  Checking a loop in "i32_factor_4"
-; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_8:          Found an estimated cost of 16 for VF 8 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp2, align 4
-; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_8-NEXT:     Found an estimated cost of 16 for VF 8 For instruction: store i32 0, i32* %tmp3, align 4
 ; VF_16-LABEL: Checking a loop in "i32_factor_4"
-; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_16:         Found an estimated cost of 32 for VF 16 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp2, align 4
-; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_16-NEXT:    Found an estimated cost of 32 for VF 16 For instruction: store i32 0, i32* %tmp3, align 4
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 1
   %tmp2 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 2
   %tmp3 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 3
   %tmp4 = load i32, i32* %tmp0, align 4
   %tmp5 = load i32, i32* %tmp1, align 4

 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load half, half* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load half, half* %tmp2, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load half, half* %tmp3, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store half 0xH0000, half* %tmp0, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store half 0xH0000, half* %tmp2, align 2
 ; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store half 0xH0000, half* %tmp3, align 2
 ; VF_8-LABEL:  Checking a loop in "f16_factor_4"
-; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load half, half* %tmp0, align 2
+; VF_8:          Found an estimated cost of 8 for VF 8 For instruction: %tmp4 = load half, half* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load half, half* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load half, half* %tmp2, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load half, half* %tmp3, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store half 0xH0000, half* %tmp0, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store half 0xH0000, half* %tmp2, align 2
-; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store half 0xH0000, half* %tmp3, align 2
+; VF_8-NEXT:     Found an estimated cost of 8 for VF 8 For instruction: store half 0xH0000, half* %tmp3, align 2
 ; VF_16-LABEL: Checking a loop in "f16_factor_4"
-; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load half, half* %tmp0, align 2
+; VF_16:         Found an estimated cost of 16 for VF 16 For instruction: %tmp4 = load half, half* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load half, half* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load half, half* %tmp2, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load half, half* %tmp3, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store half 0xH0000, half* %tmp0, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store half 0xH0000, half* %tmp1, align 2
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store half 0xH0000, half* %tmp2, align 2
-; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store half 0xH0000, half* %tmp3, align 2
+; VF_16-NEXT:    Found an estimated cost of 16 for VF 16 For instruction: store half 0xH0000, half* %tmp3, align 2
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %f16.4, %f16.4* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %f16.4, %f16.4* %data, i64 %i, i32 1
   %tmp2 = getelementptr inbounds %f16.4, %f16.4* %data, i64 %i, i32 2
   %tmp3 = getelementptr inbounds %f16.4, %f16.4* %data, i64 %i, i32 3
   %tmp4 = load half, half* %tmp0, align 2
   %tmp5 = load half, half* %tmp1, align 2

 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp5 = load float, float* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp6 = load float, float* %tmp2, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp7 = load float, float* %tmp3, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store float 0.000000e+00, float* %tmp0, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store float 0.000000e+00, float* %tmp2, align 4
 ; VF_2-NEXT:     Found an estimated cost of 24 for VF 2 For instruction: store float 0.000000e+00, float* %tmp3, align 4
 ; VF_4-LABEL:  Checking a loop in "f32_factor_4"
-; VF_4:          Found an estimated cost of 144 for VF 4 For instruction: %tmp4 = load float, float* %tmp0, align 4
+; VF_4:          Found an estimated cost of 8 for VF 4 For instruction: %tmp4 = load float, float* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load float, float* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load float, float* %tmp2, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load float, float* %tmp3, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store float 0.000000e+00, float* %tmp0, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store float 0.000000e+00, float* %tmp2, align 4
-; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store float 0.000000e+00, float* %tmp3, align 4
+; VF_4-NEXT:     Found an estimated cost of 8 for VF 4 For instruction: store float 0.000000e+00, float* %tmp3, align 4
 ; VF_8-LABEL:  Checking a loop in "f32_factor_4"
-; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load float, float* %tmp0, align 4
+; VF_8:          Found an estimated cost of 16 for VF 8 For instruction: %tmp4 = load float, float* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load float, float* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load float, float* %tmp2, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load float, float* %tmp3, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store float 0.000000e+00, float* %tmp0, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store float 0.000000e+00, float* %tmp2, align 4
-; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store float 0.000000e+00, float* %tmp3, align 4
+; VF_8-NEXT:     Found an estimated cost of 16 for VF 8 For instruction: store float 0.000000e+00, float* %tmp3, align 4
 ; VF_16-LABEL: Checking a loop in "f32_factor_4"
-; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load float, float* %tmp0, align 4
+; VF_16:         Found an estimated cost of 32 for VF 16 For instruction: %tmp4 = load float, float* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load float, float* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load float, float* %tmp2, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load float, float* %tmp3, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store float 0.000000e+00, float* %tmp0, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store float 0.000000e+00, float* %tmp1, align 4
 ; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store float 0.000000e+00, float* %tmp2, align 4
-; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store float 0.000000e+00, float* %tmp3, align 4
+; VF_16-NEXT:    Found an estimated cost of 32 for VF 16 For instruction: store float 0.000000e+00, float* %tmp3, align 4
 for.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
   %tmp0 = getelementptr inbounds %f32.4, %f32.4* %data, i64 %i, i32 0
   %tmp1 = getelementptr inbounds %f32.4, %f32.4* %data, i64 %i, i32 1
   %tmp2 = getelementptr inbounds %f32.4, %f32.4* %data, i64 %i, i32 2
   %tmp3 = getelementptr inbounds %f32.4, %f32.4* %data, i64 %i, i32 3
   %tmp4 = load float, float* %tmp0, align 4
   %tmp5 = load float, float* %tmp1, align 4